Skip to content

feat(skills): atomically promote safe auto-skill updates - #18595

Closed
rubencu wants to merge 2 commits into
kirodotdev:mainfrom
rubencu:feat/atomic-skill-candidate-claim-skills
Closed

rubencu wants to merge 2 commits into
kirodotdev:mainfrom
rubencu:feat/atomic-skill-candidate-claim-skills

Conversation

@rubencu

@rubencu rubencu commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Prose-only updates to existing auto-skills can be applied without review when the operator opts in, and every candidate promotion consumes or restores exactly the authenticated generation it inspected.

With skills.approval_required=false, safe prose-only updates to existing auto-skills still required manual approval, and there was no way to opt into applying them unattended. Candidate promotion also mutated private claim state before proving the private root was agent-denied, while rejection restoration reserialized metadata and changed CRLF bytes on Windows.

Why it matters

Users who opt out of approval and also opt into unattended updates should receive safe prose improvements without granting unattended execution to script-bearing candidates; turning approval off alone must not widen what it already meant. Promotion and recovery must consume or restore exactly the authenticated generation, never mutate an untrusted root first, inspect a public replacement, or normalize rejected candidate bytes.

Not a goal

  • Absorbing feat(skills): audit pending and live overlap (#11311) #11323 (queue-wide audit, restage, in-process target mutation lock). The two touch the same pending-update functions; the overlap is listed under Design questions so they can be sequenced, not merged here.
  • Changing the NEW-candidate policy fix(skills): respect disabled approval setting #7617 settled. This change applies the same all-invalid-script rule to the UPDATE path and leaves the new-candidate branch as it is.
  • Settling the supported platform contract for hosts whose agents run in a sandbox Kiro Crew does not build (native Windows, macOS with the Kiro CLI internal sandbox). This change ships the fail-closed choice as the default, pending maintainer confirmation.
  • A dashboard control for the new skills.auto_apply_updates setting. It is set with kirocrew config set.

What changed (motivation → approach → change)

Goal: let an operator who disabled review receive prose-only skill improvements, without letting an agent approve its own candidate and without widening what approval_required=false already meant.

Approach: every promotion first captures the public pending generation through authenticated no-follow reads, records its digest in a held claim lock, renames the public directory into a hidden public quarantine, and publishes only from a fresh private materialization of the captured bytes. The private authority lives at tag-grants/auto-skill-private, beneath the tag-grants directory Kiro Crew masks on Linux and macOS, and it is provisioned or certified only while that protecting mask actually applies. Every authority-backed use rechecks the live sandbox mode and revokes the process certificate with sandbox_off if the mode changes to off. The alternative this replaces, a dedicated deny leaf with its own mask, needed spawn-time materialization and verification, which turned an auto-skill problem into a refused agent spawn.

  • Unattended updates, separate opt-in. skills.auto_apply_updates (default false) applies a prose-only update when approval_required=false and auto_refine_on_deviation=false. Turning approval off alone still stages every update. A successful apply keeps the previous version under .versions/, runs the lifecycle pass with the target exempt, and raises a Skill auto-updated notification.
  • All-invalid-script rule on updates. When the update would be applied without review (approval_required=false, auto_apply_updates=true, refinement off), an update whose supplied scripts all failed validation is rejected with all_scripts_rejected; whenever update review is still on, including approval off with auto_apply_updates off, it is staged as prose for review as before (history_consolidation._stage_skill_update), as fix(skills): respect disabled approval setting #7617 does for new candidates.
  • Claim and whole-tree publication (skills_auto_pinning.py, skills_auto_authority.py, skills_auto_locks.py, skills_auto_snapshot.py, skills_auto_claims.py, and skills_auto_promotion.py, bound through the skills.py facade): capture, quarantine, private materialization, prepared journal, no-replace publication of the complete live tree, exact-inode refusal restore, and restart recovery that consumes, re-queues, or retains each claim by exact digest and tagged native identity.
  • Owner modules. skill_runtime/auto_skills.stage_skill_candidate keeps the one-slug-space allocator from fix(skills): coordinate live and pending auto-skill slug claims #12970 and now writes the candidate beneath the pinned pending parent; create_auto_skill and restore_auto_skill take the target lock inside the slug claim lock; skill_runtime/versions picks the snapshot number from the captured live generation. The facade members for create/update/delete/pin/inject/archive take the per-target lock.
  • Approval reasons kept. The *_checked variants keep main's PendingApprovalRefused vocabulary: advisory read-only prechecks refuse early with the reason, and the claim path notes reasons (stale_base, target_missing, live_exists, kind_mismatch, invalid_layout, script_validation_failed with its report) for refusals that happen after the claim. promotion_disabled is new.
  • Authority follows the live mask. Provisioning and certification require non-delegation plus credential_mask_applies(configured_sandbox_mode()). Every authority-backed entry rechecks that predicate; a live mode flip to off revokes the process binding and records sandbox_off, distinct from delegated and unavailable-mask refusals. Agent spawning remains independent of authority state.
  • Active-claim inspection fails closed. Staging reuses the private-state descriptors already pinned by pending.lock. An enumeration or identity failure aborts the entire attempt, audits active_claim_indeterminate, and retracts the detection marker for retry instead of interpreting the slug as unclaimed.
  • Committed recovery-pending updates stay lifecycle-exempt. If whole-tree publication commits but claim consumption cannot finish, the notification carries recovery_pending=true and SEL records auto_apply_committed_recovery_pending; the prepared claim journal durably exempts the live target from inactivity and max-count archival until restart recovery retires the claim. The immediate lifecycle pass now runs on this outcome, advances its throttle only on success, and a later retry reads the same persisted exemption. The exemption mirrors what restart recovery can do to a live tree: only an authenticated prepared or evidence journal naming the target exempts it. A valid journal that cannot be opened or authenticated makes every target indeterminate until its authenticated publication target can be read, because the candidate slug is untrusted and need not resemble an update target; lifecycle logs the exact journal and stopped-process restart recovery. Per-target scoping resumes only for a journal that was read and authenticated. A crash-orphaned .rename-probe-* file or other non-claim name exempts nothing. Restart recovery skips probe files and unlinks a bounded number per pass once their modification time shows them orphaned, and stale-authority inspection no longer counts them as claims in flight.
  • Standalone CLI authority, gateway parity. kirocrew consolidate and kirocrew run attempt the same certification as gateway startup, once, before constructing mutation-capable collaborators. As on the gateway, a refusal never stops the command: the initializer records the reason, every staging and promotion path refuses, memory consolidation and the task proceed, and one stderr notice says so when skills.auto_create_from_sessions is on.
  • Consolidation skips skill detection while auto-skills are off. _run_skill_detection treats auto_skill_promotion_disabled_reason() is not None as auto-skill creation off for the run: no aux model call, no candidate that could only be lost, no detection marker advanced.
  • An agent cannot lock the operator out, and an existing authority is never bypassed. A live auto/ name folds onto the canonical slug lock its first component reaches (NFKC, case folding, trailing dots and spaces stripped); a name no promotion can target under any spelling (auto/Persist_Me) is mutated by name as before the protocol instead of refused, so a planted always: true skill stays deletable, pinnable and toggleable. For canonical names, lock-free by-name mutation and dismissal are granted only when an unmasked gateway or operator process observes tag-grants/auto-skill-private absent on the selected data home, or while that process freshly re-proves the data-home-wide no-replace refusal. A process carrying the Kiro Crew sandbox marker or confirmed by the macOS kernel as Seatbelt-confined can see an empty masked tag-grants while the real authority exists, so it treats root absence and a fresh probe as indeterminate, requires the hidden target lock, and refuses mutation with the in-sandbox reason when that lock cannot be acquired. Root existence is re-read on every call. With a root present in every other state, every mutation takes the authority-backed target lock even after sandbox mode turns off; an uncertified process fails closed rather than bypassing a lock a certified process may already hold. The r3b host-wide sandbox predicates now shape only startup logging, the retirement hint, and which retirement path runs. On a mask-less host with an existing root, the single startup warning and refused mutation name kirocrew skills authority-retire: the operator stops every gateway and agent on that home, and the command takes the gateway lock and moves the authority root plus record aside under one .stale-<token> suffix. The ungated self-protection-skills-authority-retire argv floor (commit fe170cc on fix(security): deny kirocrew skills authority-retire from an agent shell #4533, landing first) denies that command through an agent shell with no catalog opt-out. Once an unmasked process observes the root absent, by-name mutation and explicit dismissal resume. The obsolete authority spellings remain first-creation checks only and are inert after the real authority exists.
  • Retirement handles the roots mask-less hosts actually carry. Where _auto_skill_sandbox_excludes_every_promoter() holds (delegated sandboxing, effective agent.sandbox off, no sandbox backend), no process on the data home can be certified, so authority-retire moves the root aside without trusting it, as the startup reseed does for a stale record. A home moved by cp -a, rsync or a backup restore (stale provenance) and a valid root whose evidence/ or public quarantine outgrew the stale-state scan both retire, and a planted always-on skill becomes removable again. Idleness comes from active claims/ alone, with one bounded scan; a claim journal there, or an overflow, refuses with the manual recovery. Wherever a promoter could exist, retirement still verifies provenance and identities first, and an unverifiable root refuses with the unchanged stopped-installation handoff.
  • Retirement holds no pin on the root it renames. A Windows directory handle opened without FILE_SHARE_DELETE forbids renaming that directory, so both retirement paths release the root pin before the renames and re-check the root by native identity at each one.
  • Retirement takes the public quarantine history with the root. A public .quarantine or .live-quarantine name is history only beside the private evidence/ entry of the same name, and evidence/ leaves with the root. _move_authority_aside therefore renames skills/auto/.quarantine/ and skills/auto/.live-quarantine/ aside under the root's .stale-<token> suffix on both retirement paths, so a fresh root provisioned after a retire no longer inherits 1,025 orphaned names that made staging and recovery refuse forever. Nothing is deleted: kirocrew skills authority-retire prints every set-aside path and how to requeue a candidate left there, and a failed rename or sync renames every entry already moved back, the record included.
  • Claim history no longer exhausts the active-claim bound. _pending_slug_claimed and _recover_abandoned_claims skip a public quarantine name whose claim already retired into private evidence/ before it counts toward _ACTIVE_CLAIM_SCAN_LIMIT. Every promotion, dismissal and TTL prune leaves one such name forever, so counting them froze staging for every slug, and stopped recovery from requeuing interrupted claims, after 1,025 lifetime claims. Names without evidence (including any planted ones) still count and still make the verdict indeterminate.
  • A refused dismissal is a 409, not a 404 (dashboard/handlers/prompts_skill_pending.py, minimal edit). dismiss_pending_skill_checked raises PendingDismissalRefused("promotion_disabled") when the candidate is still staged, by-name dismissal is not ruled in, and the process holds no authority certificate. api_skill_pending_dismiss maps it to 409 pending_dismissal_refused with detail = auto_skill_promotion_disabled_reason() and the retire hint where it applies, audited as rejected. The unsuffixed dismiss_pending_skill keeps False for every other caller.
  • A refused live delete or pin is a 409 too (dashboard/handlers/prompts_skill_edits.py, minimal edit). delete_skill_checked and set_pinned_checked raise LiveSkillMutationRefused("promotion_disabled") when the live auto/ skill is still there, its name takes the authority-backed target lock, by-name mutation is not ruled in, and the process holds no authority certificate. DELETE /api/skills/{name} and POST /api/skills/-/pin map it to 409 skill_mutation_refused with the same reason, detail and hint, audited rejected, instead of 404 not found and 400 not an auto-skill or not found. A skill that is gone keeps its 404 and 400. delete_skill and set_pinned keep returning bool; the three handler test doubles that stood in for delete_skill now stand in for delete_skill_checked.
  • Sensitive paths are not in this PR. The one sensitive-path piece of this change, the ungated self-protection-skills-authority-retire argv floor that denies kirocrew skills authority-retire through an agent shell (one matcher in security/argv_floor.py, its rule row and reason in security/denied_rules.py, its is_denied registration and facade export, its entry in the security spec's ungated-floor list, and the security package line budget raised to the measured 29,224), is commit 047556b8509e15aa99a2ba342171f593df05257d on the fork branch security/skills-authority-retire-floor, kept on fix(security): deny kirocrew skills authority-retire from an agent shell #4533 for a maintainer to land from a repository branch. This PR touches no path the Sensitive Change Review names. Merge order: that floor first, then this PR, so the retire command never ships without the floor that makes it operator-only.
  • The delegated-host gate covers the unreviewed path too. Wherever staging and promotion are off, create_auto_skill (which approval_required=false publishes through) returns None, and TTL pruning removes nothing, so a pre-upgrade candidate stays visible for a future supervised migration; an explicit dismissal still removes it.
  • Recoverable provenance does not bury an in-flight claim. A stale authority record (restored backup, cp -a/rsync host move, regenerated token_signing.key, renumbering remount) is reseeded at startup only after public .quarantine/ and .live-quarantine/ names and private locks/claims/ names correlate with stale private evidence. When the private authority root is missing, startup still bounded-scans both public quarantine directories and treats every entry as in flight because no private evidence remains to prove retirement; only empty public quarantine permits fresh provisioning. Any unmatched name, including the pending-to-quarantine crash window before private claim materialization, refuses reseed with the exact recovery steps.
  • Startup probes no-replace rename and bounds every claim namespace walk. Certification probes the rename primitive once inside the authority root and refuses on a filesystem without it, so the queue keeps by-name dismissal only while a call-time probe still proves every process lacks that primitive. Stale-authority inspection reads each relevant directory with os.scandir under _STALE_CLAIM_SCAN_LIMIT; overflow is indeterminate, names the breached bound, disables auto-skills, and does not delay gateway readiness with an unbounded walk. Staging streams only slug-matching claim identities and returns the existing retryable indeterminate result once more than _ACTIVE_CLAIM_SCAN_LIMIT unconsumed names are seen. Restart recovery scans a bounded active namespace across private claims and both public quarantines, drops names retired by private evidence before they count, and refuses the pass with the exact overflow reason instead of materializing and sorting permanent history.
  • Platform and sandbox. Opened-identity helpers, no-reparse handle-bound unlink, and platform_lock_compat.prepare_lock_file; pinned_fs.scan_tree_pinned gains entry and depth budgets; the skills parent sync uses main's atomic_write.fsync_open_dir rather than a helper of its own. sandbox.py and security/paths.py change comments only: no new deny leaf, mask, materialization, or spawn check.
  • Gateway startup. run_gateway provisions or certifies the authority once, before any loader or session, only when the protecting mask applies. A refusal is logged and the gateway keeps serving with auto-skill staging and promotion disabled. Certification stays on the readiness path deliberately, so no agent this gateway spawns runs before the obsolete-spelling and stale-state checks; its work is bounded and listed beside _STALE_CLAIM_SCAN_LIMIT (a few lstats, one provenance read and MAC, at most six bounded directory scans only for a stale record, provisioning syncs, one rename probe).
  • Round 8: real-Windows and macOS CI. (a) No product code changed: each of the twelve reds traces to a test double or harness, and none reaches a production path with a real defect. (b) test_fresh_root_swap_after_provenance_sync_refuses now carries requires_renamable_pinned_directory, a capability probe, because a Windows pin omits FILE_SHARE_DELETE and refuses the injected rename before any swap exists, and production never creates the authority on Windows. (c) The doubles now behave like the hosts they stand in for: native Windows keeps the real pin_directory and open_file_no_reparse (the CRT cannot open a directory), the native rollback test creates the roots it pins, identity mismatches stay in their own encoding (a native 128-bit id flips a bit instead of passing through int()), the stat double keeps st_file_attributes, the fallback claim stages with the host's real dir_fd capability, the staging tamper writes by public name, _WinEntry delegates the rest of os.DirEntry (is_junction), the shim opens a link through O_SYMLINK where there is no O_PATH (macOS), and the POSIX descriptor-identity test pins its branch. The macOS test_mcp_gateway_backend_record red is base-owned: this change touches nothing it imports.

Backwards compatibility

Breaking: auto-skill staging, approval, unattended apply, and unreviewed direct publication need a certified private authority protected by the live Kiro Crew credential mask, so they refuse where none exists or where the configured sandbox mode changes to off. A mode flip to off revokes the process certificate and reports sandbox_off before the next authority-backed operation. Live mutation and claim-based dismissal take the same authority lock whenever an unmasked boundary observes tag-grants/auto-skill-private present; an agent-sandboxed or Seatbelt-confined process treats the masked view as indeterminate and requires that lock even when the root appears absent. An uncertified process fails closed rather than racing a certified process. They retain by-name behavior only when an unmasked process observes that root absent or freshly re-proves the data home's no-replace refusal. The same staging refusal applies on native Windows and on macOS with the Kiro CLI internal sandbox (fail-closed default, pending maintainer confirmation), a gateway whose startup refused (a data home that, when its authority is first created, still holds skills/auto/.private or a top-level auto-skill-private from an earlier revision of this change, a data home whose filesystem has no atomic no-replace rename, or a stale authority record with a claim in flight), and a process that never runs authority initialization. A stale authority record with no claim in flight is reseeded at startup instead of refusing. On a mask-less host with an existing root, kirocrew skills authority-retire moves the root and record aside after the operator stops every gateway and agent, without trusting an unverifiable root there because no process on that home can be certified; elsewhere it verifies first. The ungated self-protection-skills-authority-retire argv floor denies that command through an agent shell with no catalog opt-out, restoring by-name mutation without deleting evidence. A dismissal the authority rule refuses while its candidate is still staged answers 409 pending_dismissal_refused instead of 404; the Skills tab renders it through its existing uncoded dismiss-failure copy. A live auto/ delete or pin the same rule refuses while the skill is still there answers 409 skill_mutation_refused instead of 404 and 400; retirement also moves the public skills/auto quarantines aside with the root. Wherever staging refuses, consolidation skips skill detection and TTL pruning removes nothing. skills.auto_apply_updates is a new key defaulting false, so an approval-off install keeps staging updates as it does today; every other key and file main accepts is still accepted. Writer sweep on origin/main @ ffd9136: the auto-skill writers are history_consolidation (stage new/update, direct publish), the dashboard pending handlers (approve, dismiss), prune_pending, and the live-mutation members (create/update/delete/pin/inject/archive/restore); each is listed below.

Reader: src/kiro_crew/history_consolidation.py:HistoryConsolidator._stage_skill_update -- chat -- without authority staging returns None and is audited creation_failed; with it, unchanged unless auto_apply_updates is on; an update whose scripts all failed is rejected only when it would be auto-applied (TestDelegatedSandboxFailsClosed, test_disabled_approval_alone_never_auto_applies_an_update, test_update_whose_scripts_all_failed_stays_reviewable_while_update_review_is_on)
Reader: src/kiro_crew/history_consolidation.py:HistoryConsolidator._run_skill_detection -- chat -- wherever staging and promotion are off, detection is skipped with no aux model call and no marker, while memory consolidation proceeds (test_consolidation_still_runs_on_a_delegated_host)
Reader: src/kiro_crew/history_consolidation.py:HistoryConsolidator._process_auto_skills -- chat -- direct publish of a NEW prose skill with approval off is off under the same gate as staging (test_unreviewed_publication_is_off_under_the_same_gate)
Reader: src/kiro_crew/dashboard/handlers/prompts_skill_pending.py:api_skill_pending_approve -- crew page -- promotion_disabled returns the coded 409 with the reason (test_approve_refusal_vocabulary_is_fully_classified)
Reader: src/kiro_crew/dashboard/handlers/prompts_skill_pending.py:api_skill_pending_dismiss -- crew page -- dismissal runs by name only when an unmasked process observes the authority root absent or freshly re-proves the no-replace refusal; an in-sandbox view treats both verdicts as indeterminate, and with an authority possible it takes the lock and fails closed if the process cannot authenticate it; a candidate still staged when that refusal happens answers 409 pending_dismissal_refused with the reason and retire hint, audited rejected, while a missing candidate keeps its 404 (test_unmasked_root_absence_still_grants_by_name_mutation, test_in_sandbox_root_absence_and_fresh_no_replace_probe_both_refuse_mutation, test_masked_authority_view_never_proves_real_root_absent, test_startup_refuses_a_filesystem_without_no_replace_rename, test_dismiss_refused_by_the_authority_rule_is_a_coded_409_not_a_404, test_dismiss)
Reader: src/kiro_crew/dashboard/handlers/prompts_skill_edits.py:api_skill_detail -- crew page -- a canonical auto/ live mutation takes the authority lock while its root exists and resumes by name after verified operator retirement; a name no promotion can target still runs by name, and an alias spelling waits on its canonical target lock; a delete that rule refuses while the skill is still there answers 409 skill_mutation_refused with the reason and retire hint, audited rejected, while a missing skill keeps its 404 (test_existing_authority_never_grants_lock_free_mutation_when_sandbox_turns_off, test_operator_retire_moves_verified_idle_authority_and_restores_by_name_mutation, test_a_planted_non_canonical_auto_skill_stays_deletable, test_an_alias_spelling_waits_on_the_canonical_target_lock, test_delete_and_pin_refused_by_the_authority_rule_are_coded_409s)
Reader: src/kiro_crew/dashboard/handlers/prompts_skill_edits.py:api_skill_pin -- crew page -- a pin the authority rule refuses while the skill is still there answers 409 skill_mutation_refused, audited rejected; a missing or non-auto skill keeps its 400 (test_delete_and_pin_refused_by_the_authority_rule_are_coded_409s, test_pin_roundtrip)
Reader: src/kiro_crew/skill_runtime/auto_skills.py:prune_pending -- chat -- prunes nothing while staging and promotion are off (test_pre_upgrade_candidates_survive_ttl_pruning)
Reader: src/kiro_crew/cli_knowledge.py:_consolidate_cmd -- release -- attempts certification once; a refusal keeps consolidation running with skill detection off and one notice when auto-skill creation is on, while the list-only form remains read-only (test_standalone_consolidation_certifies_before_staging, test_consolidation_still_runs_on_a_delegated_host, test_standalone_refusal_is_silent_when_auto_skills_are_off)
Reader: src/kiro_crew/cli_server.py:_run_task -- release -- after validating the spec exists and loading config, attempts certification once before constructing sessions, stores, lifecycle writers, or the consolidator; a refusal keeps the task running with one notice when auto-skill creation is on (test_authority_is_certified_before_task_state, test_authority_refusal_keeps_the_task_running, test_authority_refusal_is_silent_when_auto_skills_are_off)
Reader: src/kiro_crew/cli_knowledge.py:_skills_cmd -- release -- kirocrew skills authority-retire holds the gateway lock; on a host where no process can be certified it moves an unverifiable root aside once active claims/ is empty, elsewhere only verified idle authority state, retaining the exact root, record and public .quarantine/.live-quarantine directories under one stale token, so a later fresh root starts with no claim history, and holding no pin on the root it renames (test_authority_retire_cli_holds_the_gateway_lock, test_authority_retire_cli_reports_a_root_retired_without_a_record, test_operator_retire_moves_verified_idle_authority_and_restores_by_name_mutation, test_maskless_retire_sets_aside_an_unverifiable_root_without_trusting_it, test_maskless_retire_ignores_retired_history_past_the_scan_bound, test_masked_host_retire_refuses_an_unverifiable_root_with_the_exact_handoff, test_maskless_retire_refuses_a_claim_in_flight_and_names_the_manual_recovery, test_simulated_windows_retire_holds_no_pin_on_the_root_it_renames, test_retire_sets_quarantine_history_aside_so_a_fresh_root_stages_again, test_a_failed_quarantine_set_aside_restores_every_retired_entry, TestSelfProtectionFlagInterposition)
Reader: src/kiro_crew/skills.py:SkillsLoader._pending_slug_claimed -- chat -- staging counts only unconsumed public quarantine names toward the active-claim bound, so claim history never refuses a slug while planted names still make it indeterminate (test_consumed_quarantine_history_never_exhausts_the_active_claim_bound, test_unconsumed_quarantine_names_still_bound_the_active_claim_scan)
Reader: src/kiro_crew/skills.py:SkillsLoader.list_pending_skills -- crew page -- a candidate's name is now its directory name, which equals the metadata name for every candidate staging writes (test_skills_custom_fixture_drives_live_and_pending_readers)

Tests

Current validation on 8f463d5372 (two commits directly on ccb5d34645): main’s skills.py and history_consolidation.py composition roots are preserved. Auto-skill authority, locking, snapshot, claim, and promotion behavior lives in six focused owner modules, each below 2,000 lines; history auto-apply behavior lives in history_consolidation_skills.py. Main’s #20844 split cli.py into parser registrars and mirror owners, so _skills_cmd and the standalone authority certification in _consolidate_cmd live in cli_knowledge.py, the skills authority-retire parser in cli_parser_agent.register_skills_parser, and cli.py keeps the dispatch branch, the registrar call, and TYPE_CHECKING names for the two skills functions the owner imports. The skill-versioning tests remain a scoped package with every file at most 1,711 lines. platform_compat.rename_no_replace, which nothing called, is removed with its four tests, and the spec’s live-missing restore sentence names the pinned no-replace rename the code performs (_rename_skill_child_no_replace). Main’s #21347 folded the three nested _restore_redacted closures in skills_review.py into one module-level helper; this PR removes in-place candidate redaction, the helper’s only use, so skills_review.py carries no restore helper. The four test files this change formats (test_consolidate_cmd.py, test_skill_pending.py, test_skill_pending_notify.py, test_skill_update_flow.py) are black-clean, so they leave .github/black-baseline.txt; the baselined Black gate refuses a clean file that is still listed. Main’s #21438 split test_history.py by class, and this PR’s history test edits moved with their classes into test_history_auto_skills.py and test_history_consolidator.py, each of which defines the _initialize_skills_authority helper. This PR no longer changes test_history.py, which keeps its baseline row. Main’s #21416 split test_platform_compat.py by topic; this PR’s added platform tests stay in test_platform_compat.py with only hunk context moved. Main’s #21549 split dashboard/handlers/prompts.py into prompts_* owners behind a _facade_mirror facade: the refused-dismissal 409 and its PendingDismissalRefused import moved into prompts_skill_pending.py, and the refused delete and pin 409s, the LiveSkillMutationRefused import and the _live_skill_mutation_refused helper into prompts_skill_edits.py, directly above api_skill_pin; prompts.py keeps only the facade’s TYPE_CHECKING manifest lines for those three names. Main’s #21464 moved TestRunGateway into test_slack_gateway_run.py, and this PR’s three startup tests moved with it, so this PR no longer changes test_slack_gateway.py. Those five files are byte-identical to #19429’s resolution, which carries the same hunks. The PR’s test files plus test_frontmatter.py, the CLI split contract tests, every test_platform_compat_*.py and test_history_*.py topic module, the prompts handler and facade tests and main’s new test_consolidation_lesson_semantic_reroute.py passed 6,973 tests (39 skipped), and test_slack_gateway.py with its test_slack_gateway_*.py topic modules passed 1,038. On this head, baselined Black, isort, Flake8, mypy, line-cap, docs, comment-history, brand-name, subprocess-encoding, sync-I/O-in-async, harness-parity and diff pass, as do the flake-pattern and wait-discipline ratchets and the max-two-commit push guard.

Ran on head 07f5c68bac, one commit on a1974e202287.

  • test/test_skills_authority_retire_floor_link.py (4 tests): the CLI dispatches exactly skills authority-retire to _skills_cmd, and the operator hint names the same path; once the floor is on the base, is_denied and the floor predicate deny exactly that command (and not kirocrew skills or skills list), and the floor id is in the ungated set with no catalog row and a note naming the command. The two floor-bound tests skip on a base without the floor, with the merge order as the reason, so a rename on either side fails a test once both halves coexist.

  • test_skill_pending.py: main's document-cap fixture helper sizes its SKILL.md against the staged bytes. This change stages the generated content as UTF-8 bytes through the pinned new-file writer with no newline translation, so the on-disk length is the encoded length on every platform; the text-mode os.linesep expansion the helper assumed made the fixture 12 bytes short on Windows.

  • Round 8 (the first native-Windows and macOS CI run of this change, on fix(security): deny kirocrew skills authority-retire from an agent shell #4533's head 09f2d7096ee8, failed twelve tests (eleven PR-added, one base-owned); the test-side fixes landed in fix(security): deny kirocrew skills authority-retire from an agent shell #4533's eca89293ad and ran green on Windows on 4aec9b8622, the head this PR was carved from): new Linux doubles test_simulated_windows_identity_mismatch_refuses_untouched (both mismatch cases with native 128-bit ids) and test_simulated_windows_linked_dismissal_on_macos_open_flags (the shim without O_PATH), plus a _WinEntry completeness assertion. Against a clean archive of the previous head, scratch doubles reproduced five of the CI errors on Linux (int() of a 16-byte id for both mismatch cases, _WinEntry without is_junction in the walk shutil.rmtree runs on Windows, False is True under macOS open flags, a bytes id compared with an int inode), and all five pass on this head. Pinning roots that were never created reproduces the rollback test's WinError 2 as a missing skills root on Linux, and the setup step the test now runs makes the same pin succeed. Native Windows and macOS were never run locally: on 4aec9b8622 CI's Windows shards passed with these fixes, and the macOS shards' only failure was test_mcp_gateway_backend_record.py::test_reap_kills_a_real_term_ignoring_group_and_spares_a_recycled_one (PermissionError, not a file this change touches, the same runner-environment failure as on 09f2d7096ee8); CI's shards on this PR are the oracle. The gate files (test_skill_pending, test_skill_versioning, test_platform_compat*, test_security_gate_liveness, test_denied_commands_security, test_security_facade, test_security) passed 4,039 tests, 37 skipped. The owning set passed 5,552 tests, 36 skipped, with the one static/dist dashboard boot-chain failure classified below, which the unmodified head reproduces once that local build is present.

  • Round 6 (retirement sets the public quarantine history aside; refused live delete and pin): 80 collected node ids, asserted nonempty before expansion; 20/20 shuffled runs green at -n 0 and 10/10 green at -n 4 --dist loadgroup --max-worker-restart=2, each under a 600-second timeout with its own order and log. test_retire_sets_quarantine_history_aside_so_a_fresh_root_stages_again (masked and mask-less) plants consumed history past the active-claim bound in both public quarantines, retires, re-enables the sandbox and restarts certification: both quarantines are found intact under the root's token, nothing is deleted, and a fresh root stages, inspects and requeues an interrupted claim. test_a_failed_quarantine_set_aside_restores_every_retired_entry fails the last rename and finds the root, record and both quarantines back under their own names. test_delete_and_pin_refused_by_the_authority_rule_are_coded_409s pins both 409s, their SEL rows, the surviving skill and the unchanged 404 and 400. Mutation proof, each restored: leaving the quarantines in place turns the history test red on both paths and makes a fresh root refuse staging (active-claim inspection was indeterminate, verdict None) exactly as the review reproduced; dropping the rollback leaves the root missing; the handler calling the unchecked delete (404 == 409) or pin (400 == 409), or a classifier that never raises (404 == 409), turns the 409 test red. The owning, denied-command, security-facade and prompts-handler set passed 13,769 tests, 46 skipped, with the same 10 test_cli.py::TestDoctor* host-probe failures as on base.

  • Round 5b (retirement, active-claim bound, refused dismissal): 44 collected node ids, asserted nonempty before expansion; 20/20 shuffled runs green at -n 0 and 10/10 green at -n 4 --dist loadgroup --max-worker-restart=2, each under a 600-second timeout with its own order and log. test_maskless_retire_sets_aside_an_unverifiable_root_without_trusting_it (home moved, key regenerated, record missing) and test_maskless_retire_ignores_retired_history_past_the_scan_bound (1,025 evidence or public quarantine names) reproduce the reviewer's two triggers and prove a planted always-on skill becomes deletable; test_masked_host_retire_refuses_an_unverifiable_root_with_the_exact_handoff keeps the masked-host text exact; test_simulated_windows_retire_holds_no_pin_on_the_root_it_renames models the missing FILE_SHARE_DELETE on both paths; test_consumed_quarantine_history_never_exhausts_the_active_claim_bound stages and requeues past 1,025 consumed names while test_unconsumed_quarantine_names_still_bound_the_active_claim_scan keeps 1,025 planted names indeterminate; test_dismiss_refused_by_the_authority_rule_is_a_coded_409_not_a_404 pins the 409, its SEL row and the unchanged 404. Mutation proof, each restored: verify-first under exclusion turns 7 retirement tests red; reading retired history under exclusion turns both history cases red; holding the root pin across the renames turns the simulated-Windows case red on each path; counting consumed names turns the history test red in staging (None is False) and, separately, in recovery (candidate not requeued); calling the unchecked dismissal or never raising turns the 409 test red (404 == 409). The owning, denied-command, security-facade and prompts-handler set passed 13,765 tests, 46 skipped; its 10 test_cli.py::TestDoctor* failures fail identically on a pristine base worktree (host doctor probes). That set also exposed test_security_gate_liveness::test_the_package_stays_within_its_line_budget, red on the previous head and green on base, now re-pinned with its reason.

  • Round-five masked-view boundary: 6 collected node ids, asserted nonempty before expansion; 20/20 shuffled runs green at -n 0 and 10/10 green at -n 4 --dist loadgroup --max-worker-restart=2, each under a 600-second timeout with its own order and log. test_masked_authority_view_never_proves_real_root_absent covers both the launcher marker and kernel-confirmed Seatbelt state with a real authority hidden behind an empty tag-grants view; test_unmasked_root_absence_still_grants_by_name_mutation retains the operator-side r4 behavior; test_in_sandbox_root_absence_and_fresh_no_replace_probe_both_refuse_mutation proves neither masked absence nor a masked fresh probe grants mutation. Removing the root guard made both masked-view cases red (False is True); removing the fresh-probe guard made its regression red (True is False). Both guards were restored before the final runs. The owning, denied-command, and security-facade set passed 11,947 tests, 43 skipped.

  • Round-four regression set: 37 collected node ids, asserted nonempty before expansion; 20/20 shuffled runs green at -n 0 and 10/10 green at -n 4 --dist loadgroup --max-worker-restart=2, each under a 600-second timeout with its own order and log. The owning and complete denied-command/security-facade set passed 9,091 tests, 43 skipped.

  • Round-four mutation proof, each restored afterwards: restoring candidate-name inference for an unreadable journal turns test_an_unreadable_valid_claim_journal_exempts_every_lifecycle_target red (2 == 0 archived); granting lock-free mutation while the authority root exists turns test_existing_authority_never_grants_lock_free_mutation_when_sandbox_turns_off red; forcing the simulated-Windows fixture's delegated predicate to true turns test_simulated_windows_opened_identity_pins_delegation_false red.

  • Environment classifications, verified by the Opus lane against clean git archives: test_the_dashboard_boot_keeps_its_chain_hooks_and_order fails only when the dirty worktree's locally built website/dist is linked into the packaged static path; clean archives of base and head both pass. test_auto_improvement_members::test_assignments_build_real_context_from_only_their_private_store is inherited and identical on base because an agent-session basetemp under the crew scratch is classified as a sensitive path. The 10 test_workflows_private_execution errors are inherited and identical on base because this venv has pytest-asyncio 1.4.0 while the repository pins 0.20.3, and 1.4.0 supplies no event_loop fixture.

  • Standalone CLI coverage: test_standalone_consolidation_certifies_before_staging drives a real SkillsLoader through kirocrew consolidate and stages after certification; test_consolidation_still_runs_on_a_delegated_host drives the real initializer, consolidator, transcript and loader on a delegated host and proves the command completes, memory consolidation advances the transcript, no skill prompt is issued, no candidate exists, no detection marker is set, and exactly one notice is printed. kirocrew run keeps the task running on a refusal (test_authority_refusal_keeps_the_task_running), and both commands stay silent when auto-skill creation is off.

  • Round-two focused regressions (persisted lifecycle exemption, call-time cross-process fallback, quarantine/lock correlation, bounded stale scan, facade composition): 16 tests, 5/5 repeat runs green at -n 4. The owning and conftest-dependent set collected 8,042 tests: 8,000 passed, 41 skipped, with the same one inherited stale-static/dist dashboard boot-chain failure described below.

  • Round-three final regressions (missing-root public quarantine correlation, bounded claims and quarantine streaming, bounded recovery refusal, the existing I/O-indeterminate path): 7 tests, 20/20 repeat runs green at -n 0 and 10/10 green at -n 4 in shuffled argument order. The required owning and conftest-dependent set passed 7,723 tests, 42 skipped on the final rebased head.

  • Round-three mutation proof, each restored afterwards: restoring the missing-root else [] shortcut turns both test_missing_private_root_refuses_public_quarantine cases red by reseeding; restoring eager listdir in _pending_slug_claimed turns test_pending_slug_claim_scan_consumes_only_limit_plus_one red by returning false without consuming the bounded scanner; restoring eager recovery listings turns test_recovery_scan_overflow_is_bounded_and_indeterminate red without consuming or reporting the bounded public-quarantine scan.

  • Mutation proof, each restored afterwards: restoring the CLI early exit, or removing the detection gate, turns test_consolidation_still_runs_on_a_delegated_host red; refusing a name with no canonical fold turns test_a_planted_non_canonical_auto_skill_stays_deletable red; dropping the alias fold turns test_live_names_fold_onto_their_canonical_target_lock and test_an_alias_spelling_waits_on_the_canonical_target_lock red; re-adding the per-operation obsolete-root check turns test_an_obsolete_root_planted_after_startup_blocks_nothing red; removing the create_auto_skill gate or the prune gate turns test_unreviewed_publication_is_off_under_the_same_gate or test_pre_upgrade_candidates_survive_ttl_pruning red; never reseeding turns all four test_a_stale_idle_authority_is_reseeded_at_startup cases red, and reporting no live claims turns test_a_stale_authority_with_a_claim_in_flight_refuses_with_exact_recovery red; reverting the update rule turns both test_update_whose_scripts_all_failed_stays_reviewable_while_update_review_is_on cases red; dropping the retired-evidence filter or the startup probe turns test_recovery_skips_retired_evidence_without_opening_its_lock or test_startup_refuses_a_filesystem_without_no_replace_rename red.

  • Round-two mutation proof, each restored afterwards: dropping the claim-backed lifecycle exemption archives the committed live generation and turns test_recovery_pending_target_stays_lifecycle_exempt_until_claim_retires red; restoring the cached-startup-refusal shortcut turns test_startup_refusal_does_not_override_root_absence_after_mode_flips_back red; ignoring public quarantine correlation turns test_stale_reseed_refuses_unmatched_in_flight_state[public-only] red by reseeding; removing the named scan bound turns test_bounded_stale_claim_scan_consumes_only_limit_plus_one red.

  • Earlier rounds' mutation proof still holds: removing the live mask recheck made test_live_mode_flip_off_revokes_the_next_promotion fail with DID NOT RAISE PendingApprovalRefused; restoring fail-open return False made test_active_claim_inspection_io_failure_aborts_staging fail by staging auto/indeterminate-active-claim.

  • New behaviour pinned in earlier rounds: test_sandbox_off_has_a_distinct_refusal_and_never_provisions, test_live_mode_flip_off_revokes_the_next_promotion, test_gateway_certifies_authority_when_the_protecting_mask_applies, test_active_claim_inspection_io_failure_aborts_staging, test_auto_apply_reports_committed_update_when_consumption_needs_recovery, test_committed_auto_apply_with_pending_recovery_is_reported_as_live, the bounded concurrent-promotion barrier, and the three _floor_monkeypatch autouse fixtures. The composition signature pins now match the members earlier rounds changed (_promote_pending_update, _pending_slug_claimed, auto_apply_pending_update).

  • Gates: baseline-aware Black, touched non-baselined Black, full isort, full Flake8, mypy (2,199 source files), subprocess encoding, sync-I/O-in-async, harness parity, brand, comment-history, docs lint, git diff --check, added-line path/host/inclusive-language scans, and the denial differential (44 classified golden paths allowed at both refs, no regressions). After patch-id-identical rebases as main moved, push_guard --base main --require-single-on-base reports SAFE TO PUSH (single commit on base). The model half of the fork security-scope review is not runnable locally.

Design questions

  1. Spawn refusal (fork review item 1). No auto-skill path runs on a spawn. The authority sits under the pre-existing tag-grants mask, so namespace_argv and sandbox_exec_argv neither materialize nor verify it; a missing, refused or obsolete authority disables auto-skill promotion and never refuses an agent spawn. Pinned by TestAutoSkillAuthorityNeverGatesASpawn (no certificate, an obsolete nested root under a symlinked skills/, an older static mask already hiding the late authority).
  2. skills/auto/.private marking its ancestors (item 2). The leaf is no longer a deny-floor entry at all: authority moved under tag-grants, a single-component leaf already on the floor, so path_contains_sensitive() is false for <home>/skills and <home>/skills/auto. The obsolete spelling is a refusal input of the skill code only (test_auto_skill_authority_rides_the_tag_grants_fence).
  3. approval_required=false widening (item 3). Updates need the separate skills.auto_apply_updates opt-in; the existing switch keeps its meaning (test_disabled_approval_alone_never_auto_applies_an_update).
  4. Write after the final backup hash (item 4). Cleanup removes only private-authority trees, which no agent can reach. The public old-live generation under auto/.live-quarantine/<claim> is permanent evidence and never a cleanup target, so a write that lands after the last hash survives (test_a_late_write_to_the_old_live_generation_survives_publication_cleanup, plus the rollback-path tests).
  5. Delegated platforms: pending maintainer confirmation, implemented fail-closed. Where agents run in a sandbox Kiro Crew does not build, the gateway neither creates nor certifies authority; staging and promotion are off, while live edits and dismissal keep working because that refusal leaves the authority root absent (TestDelegatedSandboxFailsClosed). This round applies the 2026-09-18 option 1 ("disable auto-skill staging/promotion until authority was provisioned") consistently, as the reading of the pending platform contract this change takes, also pending maintainer confirmation: the unreviewed approval_required=false publication path (create_auto_skill) is off under the same gate and reports through the same auto_skill_promotion_disabled_reason(), so the less safe configuration no longer keeps a feature the reviewed one lost; consolidation skips skill detection rather than generating candidates nothing can stage; and TTL pruning removes nothing, so candidates queued before the upgrade stay visible for a future supervised migration instead of being deleted unreviewed.
  6. Lock order with fix(skills): coordinate live and pending auto-skill slug claims #12970. One order, documented once at skills.AUTO_SKILL_LOCK_ORDER: slug claim (.auto-slug-claim.lock) → claim → target → pending namespace. Every path takes a left-to-right subsequence (test_every_auto_skill_path_takes_its_locks_in_the_one_documented_order).
  7. Overlap with feat(skills): audit pending and live overlap (#11311) #11323 (for one coordination note, not absorbed): both change skills.py approve_pending_update_checked, preview_pending_update, stage_skill_candidate, _read_pending_meta, get_pending_skill, and the composition pins; skill_runtime/auto_skills.stage_skill_candidate, dismiss_pending_skill, update_auto_skill; skill_runtime/authoring.update_skill, set_pinned, set_inject_on_trigger; skill_runtime/versions.py; and history_consolidation._stage_skill_update. Locks: feat(skills): audit pending and live overlap (#11311) #11323's in-process _auto_skill_mutation_lock(name) (a per-name RLock) and this change's cross-process _live_auto_mutation_lock(name) → locks/target-<slug>.lock guard the same writers; whichever lands second should take one at the target position of AUTO_SKILL_LOCK_ORDER or replace it. Metadata: feat(skills): audit pending and live overlap (#11311) #11323's base_digest / _live_body_digest and this change's base_content_hash / canonical_skill_text_hash both hash the live body with CRLF folded to LF, and feat(skills): audit pending and live overlap (#11311) #11323's .versions/restage-<slug>.sha256 record would sit inside the whole-tree generation this change hashes and publishes.
  8. Link-screen baseline. The regenerated test/link_screen_sites.py lists 24 new screen-then-resolve sites (23 in skills.py, one in platform_compat.unlink_path_if_identity). Each refuses on a link rather than traversing one, and most act inside the masked private authority.
  9. Advisories carried from review, not implemented here.
    • Between its two renames publication briefly leaves auto/<slug> missing for by-name readers.
    • A crash between suppressed-notification staging and _auto_apply_staged_update leaves a candidate in the queue with no bell.
    • Every Windows host is delegated so production always refuses there and the ~300 lines of Windows identity/rename/delete code run only under the test shim.
    • The macOS kiro_internal_sandbox_enabled() is re-read per spawn while the delegation gate was startup-only. Verified covered: every authority-backed use calls require_auto_skill_private_authority, which re-reads sandbox.spawn_delegates_masking() and through it the uncached kiro_internal_sandbox_enabled(), so turning the internal sandbox on after startup revokes the certificate before the next stage or promotion (test_turning_on_a_delegated_sandbox_after_startup_revokes_the_next_promotion).
  10. Review findings deliberately left as they are.
    • Resolved: a lock refusal on dashboard delete and pin of a live auto-skill now answers 409 skill_mutation_refused through delete_skill_checked and set_pinned_checked, following the dismiss_pending_skill_checked precedent, so delete_skill(name) -> bool, its signature pin and its link-screen site are unchanged. The refusal is classified outside delete_skill, after it returned False, from the same predicates the lock decision reads; a reserved-namespace spelling is not an auto/ name there, so it still reads as not found.
    • The authority test shim lives in test/conftest.py only. No suite under src/kiro_crew/apps/builtins/*/tests constructs a SkillsLoader or reaches an auto-skill path, so none needs it.
    • After a reseed, a slug whose old claim left terminal public evidence in skills/auto/.quarantine/ stages its next candidate under a -N sibling, because that evidence can no longer be authenticated against the moved-aside authority. The existing pin that a lockless quarantine entry reserves its slug is kept.
  11. Retention policy for evidence-matched quarantine still needed. The active-claim scans now skip a public quarantine name with retired private evidence before it counts, so history no longer freezes staging or recovery, but both scans still read one name per lifetime claim, and private evidence/, public .quarantine/, public .live-quarantine/ and retained claim locks still grow without bound. The startup stale-state scan still counts evidence/ and the public quarantines, so a stale record on a masked host past 1,024 lifetime claims still refuses reseed and needs the manual recovery. A pruning policy for names whose public inode is provably quiescent is not built here. Operator retirement sets skills/auto/.quarantine/ and skills/auto/.live-quarantine/ aside under the root's own .stale-<token> suffix, so a fresh root provisioned after a retire starts with empty history and nothing is deleted; the retire output names every set-aside path. The startup reseed of a stale record (_set_aside_stale_authority) still leaves both public quarantines in place: it reseeds only below _STALE_CLAIM_SCAN_LIMIT, so it orphans at most that many names, but they then count toward the same-sized active-claim bound until retention exists, and its manual-recovery text does not ask the operator to move them.
  12. The Skills tab has no catalog copy for the dismissal 409. It renders pending_dismissal_refused through the existing uncoded dismiss-failure message; the response carries detail and hint for a follow-up catalog entry.
  13. Low-confidence review candidates, carried verbatim, not implemented.
    • retire excludes running gateways but not standalone certified processes and does not require a mask-less host (_pinned_parent_matches likely aborts the promoter)
    • lexists fails open rather than closed outside the sandbox case (r5 covers the in-sandbox case)
    • a transient boot refusal leaves a masked gateway uncertified for its lifetime with no hint (availability-only)
    • the ruled-out early return on the no-replace branch drops a prepared journal's exemption
    • Opus pair feat: webapp artifact kind + Artifact Deploy console (folded into Artifacts) #6 candidates 3-8 (line numbers as reviewed on f1841eac):
      • CANDIDATE 3 — src/kiro_crew/skills.py:967 — untrusted retire ignores a crash-window claim held only by a public quarantine name and its claim lock. Fix: print the requeue instructions on success when public quarantine is non-empty. Confidence: low — the crash window is narrow.
      • CANDIDATE 4 — src/kiro_crew/skills.py:846 — the masked-view guard treats every in-sandbox process as masked, whatever the data home. Fix: key the masked view on whether the selected home's tag-grants is under a masked prefix. Confidence: low — dev harness only.
      • CANDIDATE 5 — src/kiro_crew/skill_runtime/../skills.py:12049 — a certified process's transient claim failure still reports 404. Fix: raise a retryable refusal when the candidate is still staged. Confidence: low.
      • CANDIDATE 6 — src/kiro_crew/skills.py:~910 — a partial retire leaves the record retired while the root is restored. Fix: also roll the record back when its rename committed. Confidence: low — a dir-fsync failure is rare, and the state is recoverable.
      • CANDIDATE 7 — src/kiro_crew/skills.py:1361 — the retire path is chosen from the operator process's own backend probe. Fix: none clean. Confidence: low — not reproduced.
      • CANDIDATE 8 — src/kiro_crew/skills.py:9752 — the active-namespace scans now walk all history. Fix: retention (Design question 11). Confidence: low.
      • Where round 6 touches them: for candidate 3 the crash-window quarantine entry now moves aside together with its claim lock in the retired root, and the retire output prints the requeue step for a candidate left in the set-aside quarantine; for candidate 6 the generalized rollback renames every entry already moved back, the record included (test_a_failed_quarantine_set_aside_restores_every_retired_entry). Candidates 4, 5, 7 and 8 are unchanged.

Manual verification

N/A — no rendered UI changes; deterministic tests cover concurrency, restart recovery, notification, platform and sandbox behaviour. Native Windows behaviour (FileIdInfo, MoveFileW no-replace, msvcrt lock seeding) is exercised by simulation here and needs the Windows CI leg on the replacement head.

Related Issues

Fixes #3795

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

@rubencu
rubencu requested review from a team and KiroCrewCaptain as code owners October 9, 2026 20:32
@rubencu
rubencu requested a review from cruisercohen October 9, 2026 20:32
@github-actions github-actions Bot added readiness: checking Automated validation is still running fork Pull request from a fork (external contributor) readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 9, 2026
@github-actions

github-actions Bot commented Oct 9, 2026

Copy link
Copy Markdown
Contributor

GPT 6.1 Review (fork) — ⚠️ review incomplete

Reviewed 79f63912524425063450a7e048436d202cd0ba7c via the fork AI-review pipeline; updated in place on each push.

No completed GPT verdict for this commit; see the Fork GPT 6.1 Review job logs.

@github-actions

github-actions Bot commented Oct 9, 2026

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5, fork) — ⚠️ could not complete

The first-principles review did not produce a verdict for 79f63912524425063450a7e048436d202cd0ba7c (the review step never ran, because an earlier step in this job failed — no model call was made). See the Fork First Principles Review job logs. Advisory — does not block merge.

@github-actions

github-actions Bot commented Oct 9, 2026

Copy link
Copy Markdown
Contributor

Opus 5.5 Review (fork) — ⚠️ review incomplete

Reviewed 79f63912524425063450a7e048436d202cd0ba7c via the fork AI-review pipeline; updated in place on each push.

No completed Opus verdict for this commit; see the Fork Opus 5.5 Review job logs.

@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5, fork) — ⏭️ skipped

Revision 8f463d5372a36d74461a929d50d733dfc748cd6b touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Advisory — does not block merge.

@github-actions

github-actions Bot commented Oct 9, 2026

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5, fork) — ⚠️ could not complete

The design review did not produce a verdict for 79f63912524425063450a7e048436d202cd0ba7c (the review step never ran, because an earlier step in this job failed — no model call was made). See the Fork Design Review job logs. Advisory — does not block merge.

@github-actions

github-actions Bot commented Oct 9, 2026

Copy link
Copy Markdown
Contributor

Security Scope Review (fork) — 🔴 BLOCK (blocking)

Which legitimate operations does 79f63912524425063450a7e048436d202cd0ba7c newly refuse? Model proposes, scripts/deny_diff.py decides — updated in place on each push. A script-confirmed regression blocks PR readiness.

no leg reported, and this was not the nothing-to-adjudicate short circuit

No [SCOPE-REVIEWED] marker for this head was produced, so the model's text is withheld: it cannot be shown to describe this revision. See the job logs.

@rubencu
rubencu force-pushed the feat/atomic-skill-candidate-claim-skills branch from 79f6391 to 1a801c3 Compare October 10, 2026 01:10
@rubencu

rubencu commented Oct 10, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto adfd71b8, the same base the floor (#4533, 230555cd) now sits on. Same patch: stable patch-id 937edc331038 before and after, 46 files. The 21 test files this PR touches plus the security liveness, denied-command and facade suites pass locally (5,555). The six reds on 79f63912 were the container-registry rate-limit window on the first run (both image builds, Semgrep, and the four review lanes that never reached a model call, which is why Security Scope showed BLOCK without a [SCOPE-REVIEWED] leg); this push re-runs them. Merge order unchanged: the floor first, then this PR.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 10, 2026
@rubencu
rubencu force-pushed the feat/atomic-skill-candidate-claim-skills branch from 1a801c3 to 9ced455 Compare October 10, 2026 19:40
@rubencu

rubencu commented Oct 10, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main at bf1723d3f4e2: 1a801c3a -> 9ced4554c0 (one commit, same patch outside test/test_skill_pending.py, where main appended its document-cap tests and this PR appends two publication tests; both blocks are kept, main's first). On 9ced4554c0: the 23 test files this PR touches plus test_security_gate_liveness, test_denied_commands_security, test_security_facade and the ratchet modules, 5,730 passed, 36 skipped; flake8, isort, docs lint and git diff --check clean; black's seven baselined files are the same seven on main.

The floor this PR depends on is now 41d95016776087410f6ae3da9b0c1cd0acc9b1b9 on #4533 (the dependency bullet names it). The next revision here adds the test the Design lane on #4533 asked for: the registered skills authority-retire CLI subcommand path equals the floor's ("skills", "authority-retire") tuple, so a rename on either side fails a test instead of going stale silently.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 10, 2026
@rubencu
rubencu force-pushed the feat/atomic-skill-candidate-claim-skills branch from 9ced455 to 795c96d Compare October 10, 2026 20:51
@rubencu

rubencu commented Oct 10, 2026

Copy link
Copy Markdown
Contributor Author

9ced4554c0 -> 795c96d761 on main at ec2b337248 (one commit, lease 9ced4554c0), two additions and the rebase:

  • test/test_skills_authority_retire_floor_link.py, the test the Design lane on fix(security): deny kirocrew skills authority-retire from an agent shell #4533 asked for: the CLI dispatches exactly skills authority-retire to _skills_cmd, the operator hint names the same path, and, once the floor is on the base, is_denied and the floor predicate deny exactly that command while kirocrew skills and skills list stay allowed, with the floor id in the ungated set and no catalog row. The two floor-bound tests skip on a base without the floor and say why (the floor lands first); on a scratch tree carrying the fe170cc32a floor all four pass, and renaming either the floor's tuple or the CLI subcommand fails the matching test by name.
  • Backend Tests (Windows) (2) on 9ced4554c0 was test_detail_serves_a_generated_document_over_the_script_cap (6701 == 6713): main's new document-cap fixture helper sizes its SKILL.md assuming a text-mode write (\n -> os.linesep), and this change stages the generated content as UTF-8 bytes through the pinned new-file writer with no newline translation. The helper now sizes against the encoded bytes, which is what lands on disk on every platform; the Windows shard is the oracle for it.

The other reds on 9ced4554c0 were not this diff's: Dependency License Gate ("Dependency review is not supported on this repository"), Sensitive Change Review ("Could not read the PR from the API"), Frontend Tests (2) (the GuideLayer guide test, which fails intermittently across fork heads tonight and passes on main at the same base), and the fork AI lanes failing closed on the 1 MB diff as already accepted. On 795c96d761: the 24 test files this PR touches plus test_security_gate_liveness, test_denied_commands_security, test_security_facade and the ratchets, 5,732 passed, 38 skipped; flake8, isort, docs lint, comment-history gate and git diff --check clean.

@github-actions github-actions Bot added the readiness: action required A blocking check or review needs attention label Oct 15, 2026
@rubencu
rubencu force-pushed the feat/atomic-skill-candidate-claim-skills branch from aed119f to 28cac77 Compare October 15, 2026 07:36
@rubencu

rubencu commented Oct 15, 2026

Copy link
Copy Markdown
Contributor Author

Fixed the split-induced all-shard collection failure on 28cac77b21: test/skill_versioning is now a Python package, so its scoped conftest no longer shadows the repository root conftest, and all topic modules use package-relative support imports. Reproduced the cross-directory failure before the fix; afterward 453 tests passed with 2 platform skips, line-cap zero failures, and Black/isort/Flake8 passed. Max-2 pushguard and exact lease passed.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 15, 2026
@rubencu

rubencu commented Oct 16, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto 780af13e7e as d67fc72d32, preserving main’s skills.py and history-consolidation composition roots. Authority, lock, snapshot, claim, and promotion behavior now lives in six owner modules (all <2,000 lines); link-screen ownership was regenerated. Broad 2,185-test validation plus later 525/436 owner checks passed; static/docs/line-cap/max-2 guards pass. Exact lease pushed; CI is re-running.

@rubencu
rubencu force-pushed the feat/atomic-skill-candidate-claim-skills branch from 28cac77 to d67fc72 Compare October 16, 2026 18:15
@github-actions github-actions Bot removed the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 16, 2026
rubencu added a commit to rubencu/KiroCrew that referenced this pull request Oct 16, 2026
…ev#3795)

Part 1 of the kirodotdev#18595 split. It carries the platform half of kirodotdev#3795: the
helpers the atomic skill candidate claim and promotion path publishes
and verifies with, and nothing that uses them yet.

platform_compat gains rename_no_replace, a path-taking adapter over the
native no-replace rename that fails closed with ENOTSUP where the host
has no such primitive, opened_file_identity and
opened_path_identity_matches, which compare an already-open file or
directory by device and inode on POSIX and by native volume and file id
on Windows, open_path_no_reparse and unlink_path_if_identity, which
delete exactly one opened Windows entry whose native identity still
matches, and the Win32 structures those read. platform_lock_compat gains
prepare_lock_file, which seeds the one inert byte msvcrt locking needs
on an empty dedicated lock file. pinned_fs.scan_tree_pinned accepts
optional entry and depth budgets enforced while the listing is consumed.

The tests cover the identity comparison on both platforms including the
FileIdInfo and legacy-id split, the no-replace adapter's refusal of an
occupied destination, the lock-file seeding, and the descriptor-bound
directory sync ignoring a substituted path. The link-screen baseline
declares the one new screen-then-resolve site, and the platform-compat
spec documents the two identity helpers.
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 16, 2026
rubencu added a commit to rubencu/KiroCrew that referenced this pull request Oct 16, 2026
…ev#3795)

Part 1 of the kirodotdev#18595 split. It carries the platform half of kirodotdev#3795: the
helpers the atomic skill candidate claim and promotion path publishes
and verifies with, and nothing that uses them yet.

platform_compat gains rename_no_replace, a path-taking adapter over the
native no-replace rename that fails closed with ENOTSUP where the host
has no such primitive, opened_file_identity and
opened_path_identity_matches, which compare an already-open file or
directory by device and inode on POSIX and by native volume and file id
on Windows, open_path_no_reparse and unlink_path_if_identity, which
delete exactly one opened Windows entry whose native identity still
matches, and the Win32 structures those read. platform_lock_compat gains
prepare_lock_file, which seeds the one inert byte msvcrt locking needs
on an empty dedicated lock file. pinned_fs.scan_tree_pinned accepts
optional entry and depth budgets enforced while the listing is consumed.

The tests cover the identity comparison on both platforms including the
FileIdInfo and legacy-id split, the no-replace adapter's refusal of an
occupied destination, the lock-file seeding, and the descriptor-bound
directory sync ignoring a substituted path. The link-screen baseline
declares the one new screen-then-resolve site, and the platform-compat
spec documents the two identity helpers.
rubencu added a commit to rubencu/KiroCrew that referenced this pull request Oct 16, 2026
…ev#3795)

Part 1 of the kirodotdev#18595 split. It carries the platform half of kirodotdev#3795: the
helpers the atomic skill candidate claim and promotion path publishes
and verifies with, and nothing that uses them yet.

platform_compat gains rename_no_replace, a path-taking adapter over the
native no-replace rename that fails closed with ENOTSUP where the host
has no such primitive, opened_file_identity and
opened_path_identity_matches, which compare an already-open file or
directory by device and inode on POSIX and by native volume and file id
on Windows, open_path_no_reparse and unlink_path_if_identity, which
delete exactly one opened Windows entry whose native identity still
matches, and the Win32 structures those read. platform_lock_compat gains
prepare_lock_file, which seeds the one inert byte msvcrt locking needs
on an empty dedicated lock file. pinned_fs.scan_tree_pinned accepts
optional entry and depth budgets enforced while the listing is consumed.

The tests cover the identity comparison on both platforms including the
FileIdInfo and legacy-id split, the no-replace adapter's refusal of an
occupied destination, the lock-file seeding, and the descriptor-bound
directory sync ignoring a substituted path. The link-screen baseline
declares the one new screen-then-resolve site, and the platform-compat
spec documents the two identity helpers.
rubencu added a commit to rubencu/KiroCrew that referenced this pull request Oct 16, 2026
…ev#3795)

Part 2 of the kirodotdev#18595 split. It carries the candidate claim and
promotion path of kirodotdev#3795: every promotion, dismissal and unattended apply
first captures the pending generation through authenticated no-follow
reads, records its digest in a held claim lock, renames the public
directory into a hidden quarantine and publishes only from a fresh
private materialization of the captured bytes. Updates publish the
complete live tree with no-replace renames and keep the replaced
generation as permanent public evidence. Restart recovery consumes,
re-queues or retains each claim by exact digest and native identity.

The private authority lives at tag-grants/auto-skill-private, beneath a
directory every Kiro Crew sandbox already masked, so no new deny leaf,
mask or spawn check is added. All auto-skill locks follow one documented
order, skills.AUTO_SKILL_LOCK_ORDER: slug claim, claim, target,
namespace. skill_runtime/auto_skills stages the candidate beneath the
pinned pending parent and takes the target lock inside the slug claim
lock, skill_runtime/versions numbers the snapshot from the captured live
generation, and constants names the authority directories.

Nothing provisions the authority yet: gateway startup, the standalone
CLI commands and the dashboard, Slack and consolidation surfaces arrive
in part 3. The test conftest models that startup lazily so every
existing suite keeps its behaviour.
rubencu added a commit to rubencu/KiroCrew that referenced this pull request Oct 16, 2026
…tdev#3795)

Part 3 of the kirodotdev#18595 split. It wires the surfaces of kirodotdev#3795 onto the
claim and promotion path part 2 shipped.

Gateway startup provisions or certifies the auto-skill private authority
once, before any loader or session, and keeps serving with auto-skill
staging and promotion disabled when that refuses. kirocrew consolidate
and kirocrew run attempt the same certification, and kirocrew skills
authority-retire moves an idle authority aside under the gateway lock.

A prose-only update to an existing auto-skill is applied without review
when the operator opts in with skills.auto_apply_updates on top of
approval_required=false. Disabling approval alone still stages every
update. An update whose supplied scripts all failed validation is
rejected when it would be applied unattended and staged as prose when
update review is on. Consolidation skips skill detection wherever
staging and promotion are off. The dashboard raises a Skill auto-updated
notification for an unattended apply, and a dismissal, delete or pin the
authority rule refuses answers a coded 409 instead of a 404 or 400.
rubencu added a commit to rubencu/KiroCrew that referenced this pull request Oct 16, 2026
…dev#3795)

Part 4 of the kirodotdev#18595 split. It carries the spec text of kirodotdev#3795 for the
code parts 2 and 3 shipped.

memory-skills-hooks.md documents the staged-approval contract: the
separate skills.auto_apply_updates opt-in, the recovery-pending outcome
and its lifecycle exemption, the bounded claim scans, the rename-probe
cleanup, the canonical text hashing at staging and apply, and the
all-invalid-script rule applied to updates. cli.md documents what
kirocrew consolidate and kirocrew run do with the authority
certification and adds the kirocrew skills authority-retire row.
security.md records that the auto-skill publication authority lives
beneath the already masked tag-grants directory, adds no deny leaf or
mask of its own, and never refuses a spawn.
@rubencu

rubencu commented Oct 16, 2026

Copy link
Copy Markdown
Contributor Author

Refreshed onto 9037cdeea5 as d90abfd8bd, preserving main’s Slack facade while retaining authority startup in the process-entry owner. The refreshed owner/composition set passed 841 tests; Black/isort/Flake8, line-cap, diff, and max-2 guards pass. Exact lease pushed; CI is re-running.

@rubencu

rubencu commented Oct 16, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto 27ab9383e2 as f00b7d255a (still two commits). The only conflict was skills_review.py against #21347, which folded the three nested _restore_redacted closures into one module-level helper. This PR removes in-place candidate redaction, the helper's only use, so the resolution keeps this PR's side: the file is byte-identical to the previous head's, and every other file's patch is identical. The PR's test files plus test_frontmatter.py (4,631 passed, 37 skipped) and test_slack_gateway.py (389) pass; Black, isort, Flake8, mypy, line-cap, docs, comment-history, brand-name, subprocess-encoding, diff and the max-two-commit push guard pass. The body's validation paragraph names the new head.

@rubencu

rubencu commented Oct 16, 2026

Copy link
Copy Markdown
Contributor Author

Now a312db565e, still two commits, directly on eb2dad9bed. Two changes since the last note:

  • platform_compat.rename_no_replace and its four tests are removed, matching feat(platform): identity and unlink helpers (part 1/4 of #3795) #19410: nothing in this PR called it. The spec's live-missing restore sentence now names _rename_skill_child_no_replace, the pinned no-replace rename the code performs.
  • The second commit removes five entries from .github/black-baseline.txt: test_consolidate_cmd.py, test_history.py, test_skill_pending.py, test_skill_pending_notify.py and test_skill_update_flow.py. They are black-clean here but not on main, and the baselined Black gate in Backend Lint refuses a clean file that is still listed.

The PR's test files plus test_frontmatter.py (4,627 passed, 37 skipped) and test_slack_gateway.py (389) pass, and so do the baselined Black, isort, Flake8, mypy, line-cap, docs, comment-history, brand-name, subprocess-encoding, sync-I/O-in-async, harness-parity and diff gates and the max-two-commit push guard. The body's validation paragraph names the new head.

@rubencu

rubencu commented Oct 16, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto 9fba246d0c as b92bddd2e8 (still two commits). The only conflict was main's #20844 cli.py split, and the CLI change moved with it exactly as on #19429: _skills_cmd and the standalone authority certification in _consolidate_cmd live in cli_knowledge.py, the skills authority-retire parser is cli_parser_agent.register_skills_parser, and cli.py keeps the dispatch branch, the registrar call and TYPE_CHECKING names for the two skills functions. Behaviour is unchanged. The PR's test files plus the CLI split contract tests (4,875 passed, 37 skipped) and test_slack_gateway.py (389) pass, and so do every script gate, mypy and the max-two-commit push guard. The Reader: lines and validation paragraph name the new owners and head.

@rubencu

rubencu commented Oct 17, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto 5de2963627 as bf073bc72e (still two commits). Main's #21416 split test_platform_compat.py into topic modules. This PR's added platform tests stay in test_platform_compat.py, which is now main's file plus exactly this PR's added lines; only hunk context moved. The PR's test files plus test_frontmatter.py, the CLI split contract tests and every test_platform_compat_*.py topic module (5,216 passed, 38 skipped) and test_slack_gateway.py (389) pass, and so do every script gate, mypy and the max-two-commit push guard. The body names the new head.

@rubencu

rubencu commented Oct 17, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto 23affa7e78. Main's #21438 split test_history.py by class, so both commits conflicted there.

  • First commit: the history test edits moved with their classes into test_history_auto_skills.py and test_history_consolidator.py, each of which defines _initialize_skills_authority. The resolved files are byte-identical to feat(skills): approval surfaces for atomic claims (part 3/4 of #3795) #19429's. Checked definition by definition against main's split copies after Black normalization (54 definitions, 0 mismatches).
  • Second commit: its test_history.py hunks only reformatted the file (no change after py310 Black normalization), and main's split already formats the moved code. So this PR no longer changes test_history.py. Main's copy is still not Black-clean, so it keeps its row in .github/black-baseline.txt; the other four formatted files still leave the baseline.
  • Head fbee1ca4828093396e7a8eda7c5280c889bc4cae (was bf073bc72e). Local, on the patch-identical replay on 1cb15f6e0f: the owning set with every test_history_*.py and test_platform_compat_*.py topic module (6,242 passed, 38 skipped), test_slack_gateway.py (285), black, isort, flake8, mypy, the line cap and docs lint. 23affa7e78 only splits four unrelated test files. On this head, black, the flake-pattern and wait-discipline ratchets and the push guard (at most two commits ahead) pass.

The fork workflows still wait for a maintainer to approve them.

A prose-only update to an existing auto-skill can now be applied without
review when the operator opts in with skills.auto_apply_updates on top
of approval_required=false; disabling approval alone still stages every
update. An update whose supplied scripts all failed validation follows
the kirodotdev#7617 rule: rejected when approval is off, staged as prose when it
is on.

Every promotion, dismissal and unattended apply first captures the
pending generation through authenticated no-follow reads, records its
digest in a held claim lock, renames the public directory into a hidden
quarantine and publishes only from a fresh private materialization of
the captured bytes. Updates publish the complete live tree with
no-replace renames and keep the replaced generation as permanent public
evidence; restart recovery consumes, re-queues or retains each claim by
exact digest and native identity.

The private authority lives at tag-grants/auto-skill-private, beneath a
directory every Kiro Crew sandbox already masked, so no new deny leaf,
mask or spawn check is added and an authority problem never refuses an
agent spawn. Gateway startup provisions it; a refusal only disables
auto-skill staging and promotion. Hosts whose agents run in a sandbox
Kiro Crew does not build never get an authority (fail-closed default).

All auto-skill locks follow one documented order,
skills.AUTO_SKILL_LOCK_ORDER: slug claim, claim, target, namespace.

Fixes kirodotdev#3795
@rubencu

rubencu commented Oct 17, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main ccb5d34645. The head is now 8f463d5372, still two commits. The same two main-side splits as on #19429 conflicted with the first commit, and this PR carries the same hunks, so the resolution is the same and the five files are byte-identical to #19429's:

  • refactor(dashboard): split prompts.py handlers into prompts_* owners #21549 split dashboard/handlers/prompts.py into prompts_* owners behind a _facade_mirror facade.
    • prompts_skill_pending.py now holds the refused-dismissal 409 and the PendingDismissalRefused import.
    • prompts_skill_edits.py now holds the refused delete and pin 409s, the LiveSkillMutationRefused import, and _live_skill_mutation_refused, directly above api_skill_pin.
    • In prompts.py, only the facade's TYPE_CHECKING manifests changed: they now name those three.
  • test: split test_slack_gateway.py into topic files, part 2 of 2 #21464 moved TestRunGateway into test_slack_gateway_run.py. The three startup tests moved with it, so that module now imports logging and gateway as gw.

Checks:

  • The added and removed lines equal the commit's own, plus those import and manifest lines.
  • Every other file's diff is unchanged, in both commits.
  • Tests: the PR's tests, the history, CLI-split and prompts tests, and main's new lesson-reroute test pass (6,973 passed, 39 skipped). The Slack gateway modules pass too (1,038).
  • The script gates, both ratchets and push_guard (--max-ahead 2) pass.

@iamwhatever

Copy link
Copy Markdown
Contributor

Thanks for this work, @rubencu. Closing this whole-change reference PR, since it has been split into the stacked series for #3795, which is where review continues:

The branch and history stay available here, and this can be reopened if the reference is needed again.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

fork Pull request from a fork (external contributor) merge conflict Branch has merge conflicts with its base — author must resolve before merge

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Auto-apply prose skill updates: atomic candidate claim (design history from PR #2713)

3 participants