Repository navigation
feat(apps): pin agent-worlds and ops-mission-control for adoption - #38443
Conversation
UX Review (Opus 5.5) — ⏭️ skippedRevision |
|
Catalog rows: kirodotdev/KiroCrewApps#86 switches agent-worlds and ops-mission-control to |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Data-only pin rows that follow the code-review-sage precedent; nothing is taken away, and a host that cannot adopt keeps running the wheel copy. Notes
[DESIGN-REVIEWED] 08b28a5 Description read[DESCRIPTION-READ] 1f35bd27a48d2bc3b3a29c4f153191cbfa7efd4648b198484c11c611e9ca6cb1 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS This is data only: two pin rows and the two version bumps they require. AGENTS.md's "No built-in apps" rule and #26967 require these apps to leave core, so it isn't a product-shape choice made inside this PR. Not checked: whether the pinned commits and hashes match the external repositories' contents. What this change shipsInventory (6 items) — 6 justifiedIntent: let hosts running the built-in Agent Worlds or Ops Mission Control switch to each app's own repository at restart, keeping their data. ADDITION.
[FIRST-PRINCIPLES-REVIEWED] 08b28a5 Description read[DESCRIPTION-READ] c80f30032e3fc4c28ad4c847afd51b1ba6da2dbfad968726a277914d9603078d That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
b43dfe0 to
b26fa70
Compare
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Approved. Review on head b26fa70 found no blocking issues. The core diff is data only: two pin rows in adoption-pins.json and in the test table, plus the two wheel version bumps that D12 requires. Reviewers recomputed each row's archive sha256 and tree digest with this head's own adoption code, and both match. pinned_tree_refusal returns none for both trees. The SDK level scan gives Agent Worlds level 8 (it is refused at 6) and Ops Mission Control level 10 (it is refused at 9). test/test_builtin_adoption_pins.py passes (94 tests), and a revert of each hunk turns a test red. Both pinned commits are the current main of their app repositories, and permissions, agentTools and cron tuples match the wheel copies.
Security: core security code does not change here. The pinned Ops Mission Control tree adds credentialMasks: ["secrets/"], which only narrows what an agent can reach. The root token file and its core mask rows stay in place as the downgrade copy, so a host that rolls back keeps its tokens fenced.
Done well: every integrity claim in the body can be reproduced, including the level just below each pin that is refused, and the Wheel-copy base commits make the port easy to audit.
Optional:
- The App repo: hygiene gate only watches src/kiro_crew/apps/builtins// and website/src/apps//. Agent Worlds' page under website/src/pages/ (WorldsPage.tsx, scenes/, useAgentSync, usePopoutSync) and src/kiro_crew/builtin_skills/ops-mission-control/ are outside it. Once this pin lands, an edit there can drift away from adopted hosts and nothing warns. A follow-up issue could let a pin row name extra wheel-copy paths.
- The body reports that the md-notebook row on main fails
verify, because #37459 counts credentialMasks as level 10. Only PR bodies record this so far. Consider an issue or a one-row PR that re-pins it at 10.
Agent Worlds and Ops Mission Control now live in their own repositories under KiroCrewCaptain, carrying every wheel-copy change on main, with green CI on ubuntu, macOS and Windows. Their rows are the output of scripts/adoption_pin.py compute and are added to adoption-pins.json and the _SHIPPED_PINS table, so a host whose record is builtin-owned moves onto the app's own tree at its next restart and keeps data/, the app secret, the grants, the crons and enabled. Agent Worlds is pinned at minAppSdkLevel 8, the level of the useHostSessions export its page calls. Ops Mission Control is pinned at 10: its tree keeps the provider tokens in data/secrets/ behind a credentialMasks directory entry, which the agent's file tools fence from that level (#37459); it copies the data-home root store in on first read and keeps it as the downgrade copy. Each wheel built-in app.json version moves to its pinned version, since a first-adoption pin must equal it (D12). Both repositories declare the same permissions as their wheel copy (Ops Mission Control adds only the mask, which narrows), so E9 holds nothing back. The owning specs record the adoption. Part of #26967
b26fa70 to
08b28a5
Compare
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
I reviewed head 08b28a5, a rebase of the previously approved b26fa70 onto current main. The range-diff shows identical added lines, and only the context row changed. The two new pins for agent-worlds and ops-mission-control are byte-identical to the earlier head. Main's code-review-sage row stays intact in both the adoption pin file and the test's shipped-pin table, with the new rows appended after it in the same order. The adoption pin test module passes on this head, and I found no behavioural, security or test defect blocking the merge.
…tories (#86) Both former built-ins switch from a builtin source to a git source in their KiroCrewCaptain repositories, pinned at the commits Kiro Crew's adoption pin table names for them in kirodotdev/KiroCrew#38443: agent-worlds 433c1e48900b (1.0.2) and ops-mission-control 92e26de80e09 (0.1.2). Each row gets the note the other adopted apps carry and the KiroCrewCaptain author URL.
Problem / Motivation
Goal: A host that has Agent Worlds or Ops Mission Control as a built-in moves onto the app's own repository at its next restart, keeping its data, its provider tokens, its settings, its crons and whether the app was enabled.
Both apps now live in their own repositories (
KiroCrewCaptain/kirocrew-app-agent-worlds,KiroCrewCaptain/kirocrew-app-ops-mission-control), carrying every change their wheel copies have onmain, with green CI on ubuntu, macOS and Windows. Their last blockers have merged: #37426 (an app page reads the person's sessions through the SDK, which Agent Worlds' page calls) and #37459 (the agent's file tools fence an installed app'scredentialMasks, which keeps Ops Mission Control's moved token store fenced as tightly as today's root file). The wheel's adoption pin table names neither app yet, so no host adopts them and every fix to either keeps having to land twice.Why it matters
Until a pin row exists, neither app can leave core: every host keeps running the wheel copy, and the removal of the built-ins in #26967 cannot start for them.
Not a goal
scripts/adoption_pin.py, or either app's declared permissions or credential masks.src/kiro_crew/apps/builtins/agent_worlds/, the Worlds page underwebsite/src/pages/,src/kiro_crew/apps/builtins/ops_mission_control/,website/src/apps/ops-mission-control/); only eachapp.jsonversion moves.ops_mission_control_secrets.jsonstays fenced, and the adopted tree keeps it as its downgrade copy. That is fix: move ops secret store under its data dir and declare its mask #34296's lane.What changed (motivation → approach → change)
Each app gets an adoption pin row, so at its next restart a host whose record is the built-in switches to the app's own repository at the pinned commit, keeping
data/,.app_secret, the grants, the crons and the enabled flag. Each row is the output ofscripts/adoption_pin.py compute, copied intoadoption-pins.jsonand the_SHIPPED_PINStable, after code-review-sage's:minAppSdkLevel433c1e48900buseHostSessions, which its page calls)92e26de80e09credentialMasksdirectory entrysecrets/)Each wheel built-in
app.jsonversion moves to its pin's: a first-adoption pin must equal it (D12).computeaccepts Agent Worlds at 7 (it reads only the manifest and Python, and finds the exactpermissions.apientries); the row says 8 because the page callsuseHostSessions(app-kit-platform §18j). Ops Mission Control's"credentialMasks": ["secrets/"]is refused below 10 since #37459.E9 holds nothing back for either app. Agent Worlds' repository declares the same
permissions.apilist as its wheel copy and nothing else. Ops Mission Control's declares the samepermissionsandagentToolsas its wheel copy, plus thesecrets/mask, which only narrows what an agent reaches. Its crons keep their names, schedules and enabled flags; their prompts name the SOPs at the adopted skill's path.Ops Mission Control's adopted tree keeps the provider tokens in
data/secrets/secrets.json, which the OS sandbox and the agent's file tools both mask. On its first read with no store there, it copies the data-home rootops_mission_control_secrets.jsonin and keeps that file, rewritten on every save or revoke, as the copy a downgraded wheel copy reads, so core's rows for it stay.What the pinned trees carry. Agent Worlds was ported from
main33979525d8; its wheel-copy changes since, #37426 (the port's basis), #37549 (no polling from a hidden tab) and #38097 (lucide chevrons), are in KiroCrewCaptain/kirocrew-app-agent-worlds#3, released as 1.0.2 by KiroCrewCaptain/kirocrew-app-agent-worlds#4. Ops Mission Control was synced tomainaf94fac51f; KiroCrewCaptain/kirocrew-app-ops-mission-control#5 carries every change since: #37218 (the ledger import stands down while a memory provider replaces Crew's memory), #38368 (Japanese strings), the comment-only #37469 and #37953, and the fixture updates #37408 and #37597 the newer host needs; KiroCrewCaptain/kirocrew-app-ops-mission-control#6 released 0.1.2. The test-home changes #38037, #38040 and #38041 serve core's temp-dir residue guard and are not ported. No later commit, up tomain758622b385, touched either wheel copy, its strings or its store art.App repo: KiroCrewCaptain/kirocrew-app-agent-worlds@433c1e4
App repo: KiroCrewCaptain/kirocrew-app-ops-mission-control@92e26de
Wheel-copy base: agent-worlds@2e2503671fed2ca0b5c76bcf158e75fe34eae0e7
Wheel-copy base: ops-mission-control@2e2503671fed2ca0b5c76bcf158e75fe34eae0e7
The owning specs record the adoption (app-kit-platform §18j,
ops-mission-control.md).Merge order: #38478 (code-review-sage), which also appended a row after md-notebook in both files, has landed; these two rows follow its row in both files.
#34296 (open, conflicting) moves the wheel copy's store to
data/secrets.json. This pin does not depend on it: an adopted host runs the repository tree, which moves the store itself. Once this lands, that change also lands in the app repository first (itsApp repo:line), matching the repository'sdata/secrets/layout, as the layout note on #34296 asks.Backwards compatibility
Compatible: two rows are added and two versions move up. A host whose record is builtin-owned adopts at the next restart and keeps
data/,.app_secret, the grants, the crons andenabled. A host that cannot adopt (an older SDK level, a pending config review, a file besideapp.json) keeps the wheel copy unchanged.Removes nothing: both wheel copies, their
/worldsand/ops-mission-controlroutes, Ops Mission Control's root token file and its core rows, and every existing pin row stay.Tests
test/test_builtin_adoption_pins.py::test_the_shipped_pin_file_validates_and_pins_exactly_the_shipped_tabletakes the two new rows and fails with the pin file ofmain;test_every_shipped_pin_matches_its_wheel_built_in_versionholds the two version bumps.Proof: PROVEN test/test_builtin_adoption_pins.py::test_the_shipped_pin_file_validates_and_pins_exactly_the_shipped_table (head 08b28a5)
Determinism: no test logic changes, only the pinned data table. Run serially on this branch: every
test/file that namesadoption_pin,adoption-pins,agent_worlds,agent-worlds,ops_mission_controlorops-mission-control(52 files,test_builtin_adoption_*.py,test_agent_worlds_permission_parity.py,test_builtin_permission_parity.py,test_core_ops_mission_control_decoupling.py,test_app_agent_tools_adopted.pyandtest_released_builtin_skill_copies.pyamong them), both wheel copies'tests/andtest/test_flake_pattern_ratchet.py: 5454 passed, 7 skipped. Black, isort, flake8, mypy, subprocess-encoding, comment-history, brand-name, line-cap and harness-parity gates andscripts/docs-lint.shpass..github/scripts/adopted_wheel_copy_check.pypasses on this body, and fails when the twoWheel-copy base:lines name the previous port bases33979525d8andaf94fac51f(namingagent_worlds/app.json, andledger_index.py,models.pyandSettingsPanel.tsx). No frontend change.Manual verification
scripts/adoption_pin.py computewith a throwawayHOMEandKIROCREW_HOME: agent-worlds at433c1e48900bwith--min-sdk-level 8prints the committed row, and 7 prints the same row at 7 while 6 is refused for the exactpermissions.apientries; ops-mission-control at92e26de80e09with--min-sdk-level 10prints the committed row, and 9 is refused for thecredentialMaskstool-gate use. Each pinned commit is the head of its repository'smain, with every check run green. Every row measured over the network before #38478 landed (one at a time, sinceverifystops at its first refusal): sixteen rowsok, both new ones included; the rebase onto it changes neither new row. md-notebook's row, unchanged here, is refused: it was pinned at 8 one hour before #37459 made anycredentialMasksa level-10 use.Upgrade smoke on isolated loopback gateways (throwaway homes, each app's row as
KIROCREW_HARNESS_ADOPTION_PINS, a real GitHub fetch). Each home was first booted onmainwith an empty harness table, so the record was the built-in, then this branch booted twice (STAGE, APPLY), once with the app enabled at the upgrade and once with it disabled.Agent Worlds (a marker file in
data/, since the app keeps no state of its own):origin: registry, commit433c1e48900b, version 1.0.2,enabledkeptdata/byte-equal, grants kept, trust bound to the repository/api/appsrow carries the legacy nav/worlds,Gamepad2ui/dist/index.mjs;/worldsand/apps/agent-worldsrender in headless Chromium with no page error/api/crons,/api/spawn,/api/chat/slots, ...) answer 200data/; after a restart no built-in comes backThe refusal of a path outside
permissions.apihappens in the browser SDK, so a gateway probe cannot see it; the app's tests andAppSdkHostSessions.test.tsxpin it.Ops Mission Control (a provider token saved through the app's own route into the root file, settings, a ledger entry, a cron moved off its default):
origin: registry, commit92e26de80e09, version 0.1.2,enabledkept, trust bound to the repository, legacy nav/ops-mission-control,Radioconfig.json,ledger.jsonl, the grant store, the policy file and the root token file unchangedledger-hygieneincluded); the SOP path their prompts name existsmain)ui/dist/index.mjs;/ops-mission-controland/apps/ops-mission-controlanswer 200GET /stateshows the provider configured; the first read copies the token intodata/secrets/secrets.json(0600 in a 0700 directory), same value by hash, root file keptcatcannot read either token file, while the ledger readsops_mission_control_apiis listed andGET /stateanswers 200 through it; a path outside its list is refuseddata/withsecrets/and the root file; after a restart no built-in comes backDisabling and enabling the adopted app re-creates its crons from the manifest, so a cron the person switched off comes back at its default; the built-in on
maindoes exactly the same, so this pin changes nothing there.No ERROR or traceback in any Ops Mission Control gateway's stderr across 14 boots. The only tracebacks in the Agent Worlds runs come from the dashboard's own ACP spawns when the headless browser opened it ("Not logged in": the throwaway homes have no CLI login), the same in a built-in baseline home. Reinstall ran against the published catalog with only the app's row switched to
gitat its pinned commit (the KiroCrewApps PR below), served atofficial_catalog._open_catalog; with today'sbuiltinrows an adopted record has no install source until that PR lands, as for every earlier pin.Related Issues
Part of #26967
Unblocked by #37426 and #37459. Catalog rows: the kirodotdev/KiroCrewApps PR linked in a comment below switches both apps to
gitat the same commits. It is not auto-merged: the driver merges it right after this PR lands.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributed under the project's Apache-2.0 license.