Skip to content

docs: add guide for custom Anthropic-compatible LLM endpoints - #8657

Merged
NicholasRBowers merged 1 commit into
kirodotdev:mainfrom
th3l4ughingm4n:docs/custom-llm-backend
Sep 24, 2026
Merged

NicholasRBowers merged 1 commit into
kirodotdev:mainfrom
th3l4ughingm4n:docs/custom-llm-backend

Conversation

@th3l4ughingm4n

@th3l4ughingm4n th3l4ughingm4n commented Sep 5, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

The claude harness is publicly selectable (ACP_BACKEND_CLAUDE is in
BASELINE_SELECTABLE_BACKENDS, and acp/client.py owns its full spawn path),
but the operator path to actually run a session against a non-Anthropic
endpoint — DeepSeek, a self-hosted gateway, any Anthropic-compatible API —
exists only in code comments and
docs/system-specs/modules/claude-code-provider.md. A user who wants it has
to reverse-engineer which binaries the harness needs, which environment
variables route the request, and where those variables must live for each way
the gateway is launched. The public docs still describe model requests as
following kiro-cli's account alone.

Why it matters

Operators with their own endpoint cannot use the harness today without reading
the source. The launch-path gotchas are the expensive part to get wrong: a
macOS GUI launch never reads a shell profile, so the standard export recipe
silently does nothing and the session falls back to the wrong credentials.
Documenting the flow turns a selectable harness into an actually usable one for
the public build.

What changed

Goal: one task-oriented guide for running the agent on the operator's own
Anthropic-compatible endpoint via the claude harness, with the security
boundaries stated exactly as the owning spec states them — and the env
passthrough that auth depends on pinned by a test, not just prose.

Piece What it does
docs/guides/custom-llm-backend.md (new) the operator recipe: harness switch, prerequisites, endpoint routing, credential delivery, models, account boundaries
docs/system-specs/modules/claude-code-provider.md one added contract sentence: the gateway forwards inherited ANTHROPIC_* / CLAUDE_CODE_* to the harness child, so the spawn scrub list must not grow to cover them
src/kiro_crew/sandbox.py a comment beside the env scrub lists pointing at that passthrough contract — the one place a contributor hardening the lists would look; no value or logic change
test/test_sandbox_argv.py a regression test pinning the passthrough contract: neither scrub list may match ANTHROPIC_* / CLAUDE_CODE_* keys, and scrub_agent_subprocess_env passes them through — a hardening edit that breaks custom-endpoint auth fails red instead of shipping green
docs/guides/README.md index entry for the new guide

The guide covers: the agent.acp_backend: "claude" switch and its
degrade-to-kiro safety net; the claude-agent-acp npm package and claude
CLI prerequisites, with the dashboard's backend probe as the live check; and
endpoint routing (ANTHROPIC_BASE_URL, ANTHROPIC_MODEL,
CLAUDE_CODE_SUBAGENT_MODEL) on each launch path, including the
launchctl setenv recipe for the macOS desktop app and its exposure note.

Credential delivery is two-path. The recommended path is Claude Code's own
apiKeyHelper in ~/.claude/settings.json: the helper's output is sent as the
auth headers and refreshed on CLAUDE_CODE_API_KEY_HELPER_TTL_MS, which keeps
the raw token out of the ambient environment that every agent-run child
inherits. The guide also states what this does not protect: code running as the
same user can still reach the helper and its store, and the helper itself runs
inside Crew's subprocess sandbox — which hides the home directories the common
secret stores read from, so a helper shelling out to pass or a cloud-CLI
secrets manager comes back empty in the default sandbox. The guide steers the
reader to a source that survives (the macOS keychain, or a store kept outside
the home directory) and defers the exact masked-directory set to
_STANDARD_DIRS in sandbox.py and the spec. The fallback is
ANTHROPIC_AUTH_TOKEN in the gateway environment, kept working but carrying
the plain warning that every agent-run child inherits it and a prompt-injected
command can read it. The guide notes the MCP secrets vault cannot deliver this
token, since it injects only into bound MCP server environments.

The account-boundary sections are written as non-exhaustive statements rather
than enumerated exceptions: chat and worker model turns go to the operator's
endpoint, while agent-internal and background paths can stay on the kiro-cli
sign-in (crew-member DM threads follow agent.member_acp_backend). The MCP
section keeps the security kernel — Crew authors the session's settings seed;
tool calls pre-approved in the operator's own ~/.claude settings bypass
Crew's approval gate and audit log; a governed registry-mode install may
withhold spec-referenced servers — and defers finer mount rules to the linked
spec section.

Tests

test/test_sandbox_argv.py::TestAgentEnvPassthroughContract locks in the
spec's must-not-grow clause: for five representative ANTHROPIC_* /
CLAUDE_CODE_* keys, no _SENSITIVE_ENV_PREFIXES prefix matches, no
_AGENT_DENIED_ENV_KEYS prefix matches, and scrub_agent_subprocess_env
returns them unchanged. Prefix semantics are asserted (both lists match by
startswith), so a prefix entry like "ANTHROPIC_" cannot slip past an
equality check. The full module passes (219 passed, 4 skipped). The
src/kiro_crew/sandbox.py change itself is comment-only (no value or logic
change).

Manual verification

scripts/docs-lint.sh passes (indexes and internal links verified; the new
guide's relative links resolve).

Related Issues

N/A — none.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality (TestAgentEnvPassthroughContract; docs-lint and flake8 pass)
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

@th3l4ughingm4n
th3l4ughingm4n requested review from a team and cixuuz September 5, 2026 05:06
@github-actions github-actions Bot added fork Pull request from a fork (external contributor) readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Sep 5, 2026
@github-actions

github-actions Bot commented Sep 5, 2026 •

Copy link
Copy Markdown
Contributor

GPT 5.6 Review (fork) — ✅ no blocking findings

Reviewed d31454f54a309c115f8c3935ff00ebe338ca4538 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] d31454f

@github-actions

github-actions Bot commented Sep 5, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5 Review (fork) — ✅ no blocking findings

Reviewed d31454f54a309c115f8c3935ff00ebe338ca4538 via the fork AI-review pipeline; updated in place on each push.

Review details

One advisory finding: the new guide states background-worker routing as "any non-kiro harness", which is false for the kas and codex harnesses.

FINDING — docs/guides/custom-llm-backend.md:47 — "they run on the provider-backed path for any non-kiro harness" is false for two baseline-selectable backends: _bg_runtime_backends() = ACP_BACKENDS_ACP_RUNTIME ∩ ACP_BACKENDS_SESSION_EVICTION ∩ selectable_backends() = {kiro, kas, codex} (session.py:597, agent_sdk/backends.py:1290/1453, BASELINE_SELECTABLE_BACKENDS at backends.py:484), so with agent.acp_backend: "kas" or "codex" _bg_backend_supports_runtime() (session_background.py:308) returns True and get_bg_session takes the shared multiplexed-runtime branch, not the provider-backed one the guide attributes to every non-kiro harness → Fix: state it positively — a harness outside _bg_runtime_backends() (which claude is) serves background handles over the provider-backed path.

[OPUS-REVIEWED] d31454f

@github-actions

github-actions Bot commented Sep 5, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Fable 5, fork) — 🟡 CONCERNS

Design-level review of d31454f54a309c115f8c3935ff00ebe338ca4538 via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

Passthrough pin forbids the mechanism (scrub-list growth) instead of requiring the outcome, foreclosing the fix for the token leak the guide itself documents.

Watch

The spec now says "the spawn scrub list must not grow to cover either namespace," and the test pins _SENSITIVE_ENV_PREFIXES / _AGENT_DENIED_ENV_KEYS at the list level. But the guide's fallback path concedes "the harness child and every agent-run child process inherit the raw token… A prompt-injected command can read it." The legitimate future hardening — scrub ANTHROPIC_AUTH_TOKEN from agent-influenced spawns while forwarding it to the harness child via extra_env at the ACP _spawn layer — keeps the contract's real requirement intact yet goes red against this pin, and the spec/comment tell that contributor "don't" rather than naming the sanctioned route. The contract should be stated as "inherited ANTHROPIC_*/CLAUDE_CODE_* must reach the harness child."
Clears when: the spec sentence and the test docstring (or a comment in TestAgentEnvPassthroughContract) state the harness-child-outcome requirement and explicitly name selective extra_env forwarding as the permitted way to tighten the ambient scrub.

[DESIGN-REVIEWED] d31454f

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Sep 5, 2026
@bolichen97
bolichen97 force-pushed the docs/custom-llm-backend branch from 7efef71 to 3daa557 Compare September 8, 2026 12:14
@bolichen97

Copy link
Copy Markdown
Contributor

Rebased onto main 02d7a2d0 by a maintainer as part of the 2026-09-08 open-PR audit (the branch was 361 commits behind).

Conflicts: clean rebase, no conflicts.

One fix the rebase required: both links to ../system-specs/features/claude-code-provider.md are dead on current main, where that spec moved to docs/system-specs/modules/claude-code-provider.md. Repointed both (lines 17 and 125). Nothing else in the guide changed; the diff is still the same 2 files, +126 lines.

Gates run locally: python3 scripts/docs_lint.py passes (link, index-reachability and phantom-path checks clean; remaining report-only findings all pre-exist on main and are in unrelated files). No Python or frontend files touched, so the black/isort/flake8/pytest and tsc lanes do not apply.

Please review the link repoint. A maintainer push makes the maintainer the last pusher, so under this repo's last-push rule a second approver is needed. Reply here if anything looks wrong.

@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 Sep 8, 2026
@dwu96 dwu96 added the needs-pr-triage PR scanner: awaiting automated triage label Sep 17, 2026
@NicholasRBowers NicholasRBowers added drive-to-green PR claimed by drive-to-green pipeline and removed needs-pr-triage PR scanner: awaiting automated triage labels Sep 17, 2026
@NicholasRBowers

Copy link
Copy Markdown
Contributor

🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca]: This PR has been inactive for 7+ days with failing CI. I've assessed the blockers and they appear resolvable — I'll push fixes directly to this branch as a co-author.

Assessment: Both blocking lanes name the same single defect. The §5 MCP bullet in docs/guides/custom-llm-backend.md states that a claude session starts with "no Crew MCP tools", but _claude_session_mcp_servers() fills mcpServers per spawn and docs/system-specs/modules/claude-code-provider.md §"MCP tools on a Claude session" documents exactly that — so the guide inverts its own trust boundary and the "disclosed in claude-code-provider.md" claim points at a boundary that does not exist. Fix is the rewrite both reviewers spell out: state that Crew fills mcpServers per spawn, keep the already-correct sentence about pre-approved calls skipping Crew's gate, note the no-seed exception for projects carrying their own settings.local.json, and drop the "disclosed" claim. Secondly, the merge conflict is one table row in docs/guides/README.md — main rewrote the neighbouring secrets-env.md row (#10959) and inserted ten oauth-app-registration/* rows (#9992) between the PR's two context lines, so the resolution is simply re-placing the new row. Will also fold in the Design suggestion to flag that launchctl setenv ANTHROPIC_AUTH_TOKEN leaves the raw key readable via launchctl getenv to every process in the login session. Everything else in the guide was verified accurate against the base code by the Design reviewer, Opus reported no blocking findings, and all other CI lanes are green.

If you'd prefer I don't touch this PR, add the pr-no-autofix label.

@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Sep 17, 2026
@NicholasRBowers

Copy link
Copy Markdown
Contributor

Intent: Give operators a guide for running Kiro Crew's agent against a self-owned Anthropic-compatible endpoint (DeepSeek, a self-hosted gateway) via the claude harness, with security boundaries stated accurately against the owning specs.
Not a goal: Any code or behavior change; documenting setups beyond the custom-endpoint path.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca] — drive-to-green changes at 4855dab (branch rebased onto current main):

  • §5 MCP bullet rewritten to match claude-code-provider.md: claude sessions do receive Crew's MCP servers per spawn (the control plane, plus the user-configured servers the agent's spec references, after registry filtering) whenever Crew authors the session's .claude/settings.local.json. The existing pre-approval bypass caveat is kept verbatim; the no-seed/no-array exception for a project carrying its own settings.local.json is added; the false "disclosed in claude-code-provider.md" claim is dropped.
  • macOS secret-exposure note: launchctl setenv makes the raw token readable via launchctl getenv by every process in the login session — the step now says so and points at secrets-env.md.
  • docs/guides/README.md conflict resolved: kept main's rewritten secrets-env.md row and the ten oauth-app-registration/* rows; re-inserted this PR's single new index row.
  • PR body: one line updated to match the corrected §5 (it restated the old claim).

Original guide by th3l4ughingm4n, preserved as commit author. Local gates green (docs lint, related tests, both local review lanes).

@NicholasRBowers

Copy link
Copy Markdown
Contributor

🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca]

  • fixed — §5 MCP isolation claim corrected span=da67684f6e59

Fixed at 4855dab: §5 now states Crew MCP tools are mounted per spawn when Crew authors .claude/settings.local.json (agent-referenced servers, after registry filtering), keeps the sentence that user/project allow rules pre-approve calls past Crew's deny/governance/audit gate, adds the no-seed exception, and drops the false "disclosed" claim.
Wording verified against claude-code-provider.md §"MCP tools on a Claude session" and its seed-ownership section.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca]

  • fixed — stale security-boundary disclosure span=333751b237fe

Fixed at 4855dab: the guide's §5 states claude sessions receive Crew's MCP servers per spawn, consistent with claude-code-provider.md, and the "disclosed" claim for the nonexistent boundary is dropped.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca]

  • fixed — launchctl secret exposure noted span=a028dbf21a7a

Fixed at 4855dab: the macOS step now says the raw token is readable via launchctl getenv by every process in the login session and routes readers to secrets-env.md.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Sep 23, 2026
@NicholasRBowers

Copy link
Copy Markdown
Contributor

span=b1a940ee005f

  • Kiro-account isolation claim vs crew-member DM routing

fixed in b566414: member DMs route via agent.member_acp_backend (default kas = kiro-cli sign-in), ignoring acp_backend — members.py select_provider_backend, loader.py:2569. Guide now names the exception and the key to set.
Verified against the cited code before editing; docs lint green.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

span=b1a940ee005f

  • auto model claim vs warm advertised-model cache

fixed in b566414: _write_claude_local_settings omits the model key on auto once availableModels is seeded, so auto resolves to the allowlist head, not ANTHROPIC_MODEL (acp/client.py:6489-6531). Guide now scopes the claim to the cold-cache first session.
Verified against the cited code before editing; docs lint green.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

span=b1a940ee005f

  • binary-pinning rationale vs augmented PATH

fixed in b566414: augmented_path already covers Homebrew, npm global bin and version-manager shims (env.py _EXTRA_PATH_DIRS + _node_all_bin_dirs); the resolver ladder uses it (acp/client.py:1686). Guide now says pinning is for unusual locations only.
Verified against the cited code before editing; docs lint green.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

span=b1a940ee005f

  • backend-status install-command claim

fixed in b566414: backend_install.py:244 sets the install command only when the ADAPTER is missing; a missing claude CLI reports no command. Guide sentence now carries that qualifier.
Verified against the cited code before editing; docs lint green.

@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 readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Sep 23, 2026
@NicholasRBowers

Copy link
Copy Markdown
Contributor

self-added: no
mechanism: one-line cross-reference comment beside the sandbox env scrub lists (non-behavioral)

  • fixed span=fdae7c7adb13 — spawn-scrub passthrough constraint now witnessed in code

The Watch is honored via the lighter of its own two clears-when options: a comment
beside _AGENT_DENIED_ENV_KEYS/_SENSITIVE_ENV_PREFIXES in src/kiro_crew/sandbox.py
cross-references the env-passthrough contract in claude-code-provider.md and names the
guide that depends on it, so a contributor hardening the lists sees the dependency in place.
A behavioral test was ruled disproportional for this docs-scoped PR (round-5 retrospective).
Fix SHA d93a705.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

self-added: yes
mechanism: member-DM "one exception" sentence replaced by a non-exhaustive boundary statement

  • fixed span=b1a940ee005f — guide ~line 15: "the one exception is crew-member DMs" was false

The false absolute + single enumerated exception is replaced by a non-exhaustive boundary:
chat/worker turns go to the operator endpoint; agent-internal and background paths can stay
on the kiro-cli sign-in (member DMs follow agent.member_acp_backend). llm_pool.py:315
and the review pool no longer contradict the text because it no longer claims completeness.
Round-5 retrospective ruled this the smaller replacement for the round-1 accuracy patch.
Fix SHA d93a705.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

self-added: no

  • fixed span=b1a940ee005f — guide ~line 76: apiKeyHelper runs inside Crew's subprocess sandbox

Verified on code: claude is absent from ACP_BACKENDS_INTERNAL_SANDBOX
(agent_sdk/backends.py:1173), so wrap_argv_async(..., is_kiro_cli=False) wraps the
harness child and every descendant, the helper included (sandbox.py confines descendants;
filesystem masking covers key stores like .gnupg). The recipe now says the chosen secret
source must be reachable from inside that sandbox. Fix SHA d93a705.

@NicholasRBowers

Copy link
Copy Markdown
Contributor

Rebased: main moved in files this PR also touches since head b566414 went green (2fdadc7 -> 5c23469, docs/ same-dir overlap); CI re-running on d93a705.

@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Sep 23, 2026
@github-actions

github-actions Bot commented Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5, fork) — ✅ PASS

Premise-level review of d31454f54a309c115f8c3935ff00ebe338ca4538 via the fork AI-review pipeline — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: PASS

Verify the external recipe claims — npm package name, apiKeyHelper header/TTL behavior, launchctl setenv inheritance — on a real macOS GUI launch; nothing in this repo can confirm them.

What this change ships

Inventory (5 items) — 5 justified

Intent: Let an operator run the publicly selectable claude harness against their own Anthropic-compatible endpoint without reading the source — an ADDITION (guide + a pin of existing behavior).

  1. A new operator guide explains pointing the agent at a custom endpoint — justified
  2. The guides index gains a row for it — justified
  3. The provider spec now states the gateway keeps forwarding inherited ANTHROPIC_*/CLAUDE_CODE_* to the harness child — justified
  4. A comment beside the sandbox scrub lists points hardeners at that contract — justified
  5. A new test fails red if either scrub list grows to cover those namespaces — justified

Checks run: both scrub lists on base carry no ANTHROPIC/CLAUDE_CODE entry (grep of sandbox.py:4447,4500), so the pin records real base behavior; ACP_BACKEND_CLAUDE is in BASELINE_SELECTABLE_BACKENDS (agent_sdk/backends.py:487); no existing doc covers this job (grep ANTHROPIC_BASE_URL|apiKeyHelper across docs/ and src/kiro_crew/docs/: 0 hits); every symbol and spec section the guide cites (member_acp_backend, CLAUDE_AGENT_ACP_BIN, _STANDARD_DIRS, secrets-env.md, "MCP tools on a Claude session") exists on base.

[FIRST-PRINCIPLES-REVIEWED] d31454f

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Sep 23, 2026
Document running the agent on the claude harness against the
operator's own endpoint (DeepSeek, self-hosted gateways): the
acp_backend switch, the claude-agent-acp and claude CLI
prerequisites, ANTHROPIC_* environment routing on each launch
path, model resolution, and the boundaries that stay on the Kiro
account.

Original guide written by th3l4ughingm4n; review fixes applied by
Kiro Crew (per-spawn MCP seeding wording aligned with
claude-code-provider.md, launchctl secret-exposure note, guides
index rebase).

Co-Authored-By: Claude Code <noreply@anthropic.com>
Co-authored-by: Kiro Crew <noreply@kirodotdev.github.io>
@NicholasRBowers

NicholasRBowers commented Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

self-added: yes
mechanism: apiKeyHelper reachability caveat rewritten from generic hedge to concrete example correction (no _STANDARD_DIRS enumeration)

  • fixed span=b1a940ee005f in d31454f

The apiKeyHelper paragraph no longer offers pass/cloud-SM behind a generic
"must be reachable" hedge: it now says the sandbox hides the home directories
the common stores read from, names the guide's own failing examples (pass,
gcloud secrets, az keyvault come back empty in the default sandbox), and
steers to sources that survive (macOS keychain, or material outside the home).
Round-6 retrospective ruled full masked-homes enumeration a drift liability
(mutable _STANDARD_DIRS code list in prose under-warns as it grows); the
exact mask set and strict/cc nuance are deferred to sandbox.py and the spec.
Span at 3rd round: round-5's generic property wording did not clear it because
a reader still picked a failing example; correcting the examples themselves does.

@NicholasRBowers

NicholasRBowers commented Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

self-added: yes
mechanism: passthrough-negative regression test pinning both scrub lists (TestAgentEnvPassthroughContract)

  • fixed span=62fb85f1526f in d31454f

Took the Watch's first clearing condition: test/test_sandbox_argv.py:: TestAgentEnvPassthroughContract asserts no _SENSITIVE_ENV_PREFIXES prefix
and no _AGENT_DENIED_ENV_KEYS prefix matches five representative
ANTHROPIC_*/CLAUDE_CODE_* keys, and scrub_agent_subprocess_env passes
them through end to end. Prefix semantics asserted (both lists match by
startswith), so a "ANTHROPIC_" prefix entry cannot slip past equality.
Liveness mutation-proven: appending either namespace to either list turns it red.
Round-5's comment-only pin could not fail CI — the exact failure mode this
Watch named on its 2nd round — so the witness test supersedes it; the
sandbox.py cross-reference comment stays as the discoverability anchor.

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving after an automated readiness screen against this head:

  • readiness: passed label present and the PR Readiness required status is green on this head
  • no failing, cancelled or pending required check on this head
  • mergeable with no conflict against current main
  • no live reviewer CHANGES_REQUESTED (GitHub reviewDecision is not CHANGES_REQUESTED)
  • every review-lane finding on this head is either PASS, withdrawn, disposed of by the author, or advisory-only — no unresolved blocking change request

Auto-merge (squash) is enabled; branch protection still gates the merge.

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)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants