Repository navigation
feat(planning): wayfind absorbs the wayfinder decisions and accept list - #12574
Conversation
Lane-Activity: work-resolve-issue/resolve-issue
Lane-Activity: pr-unblock/address-feedback Lane-Thread: PRRT_kwDOTCGFQM6y4xIA
…1523574219-1' into fix/12524-replacement
check-next-graph failed lint-repo on #12524: wayfind (contract) routes to /work-items:decompose, past design and plan. That edge is owner decision Q531 (#12477): a map closes to prd, decompose or work. The chain roster gains a per-edge `skip-to: <plugin:skill> <reason>` declaration, and wayfind declares the one edge to decompose. Every other stage skip still fails, which a new test proves. The workflow ladder's Decompose stage now states the same entry, so ladder and gate agree. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Vcupom3BKF6QuC5U3sRJPj
The parser description no longer calls splitlines on a possibly-None __doc__, and the test module raises ImportError when the gate's spec or loader cannot be built instead of dereferencing None. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Vcupom3BKF6QuC5U3sRJPj
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 30a3875ef4
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
claude-security-review reviewed Security review: PR #12574, incremental (b01ed71 → a63f6c8)No security issues found. I read every hunk of both files in the incremental scope. Neither file is on a trusted path. This push changes no plugin manifest, workflow, action, token request, merge-hold script or trust control. What I checked
Spend: none. Nothing in this change uses prepaid credit or subscription usage, and nothing can bill. critical-findings: 0 |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
|
claude-review reviewed Code review (incremental):
|
- wayfind work step 8: post a closing comment that states every resolved decision, then hand decompose `#<map-number>` and prd the decisions as its task description, both accepted source forms. - wayfind `## Next`: one bullet per destination. - tracker-mechanics: recount after a needs-human claim; over the cap, the lowest item numbers keep the slots and the rest release. - check-next-graph: validate the raw declaration kind before rewriting it, so `entry point:` and `skip-to x:y:` exit 2; test both. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Vcupom3BKF6QuC5U3sRJPj
Ranking by item number let two sessions both keep their claims when the lower-numbered claim landed after the other session's recount. Rank by lease comment id instead: a claimant recounts only after posting its lease, so the later claimant always sees the earlier lease and backs off. Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01Vcupom3BKF6QuC5U3sRJPj
Post-merge mutation testingKilled 6, survived 1, timeout 0, no coverage 0, invalid 0, not run 0.
Surviving mutants
The run's summary has the full report. |
Refs: #12477
Refs: #12524
Summary
Replaces lanes PR #12524 (
needs-session). It merges #12524's head (c62c93e) into current main unchanged, plus one fix for thelint-repofailure the lane could not make. #12524's own summary, conflict list with Matt's wayfinder, remaining work and not-verified items carry over as written there:/planning:wayfindtakes the owner's flow, research, shared-files and picking decisions (Q531-Q533, D108) and the Topic 10 "Wayfind" accept list, with two new settings,wayfind_research_startandwayfind_hitl_concurrency. Done-when item 2 of #12477 (the interview's size threshold) is still open, so this refs the issue and does not close it.Fix
lint-repofailed onnext-graph(job):Decision: keep the lane's wayfind →
/work-items:decomposeclosure and let the gate allow that one edge. Basis: owner decision Q531, quoted in #12477 ("Decisions (owner-approved: GUIDE D108; register Q205, Q531-Q535; all 2026-10-15)", and "The map closes to prd, decompose or work"). A closed map has already resolved the design decisions a plan would hold. The stage-skip rule is a workflow gate with no allowance for that edge, so the gate is what changes. Design and Plan stay required for every other skill, and wayfind keepsworkflow-stage: contract.scripts/check-next-graph.py: the chain roster accepts a per-edgeskip-to: <plugin:skill> <reason>declaration, and the stage-skip check passes only the declared edge. A malformedskip-to(no target) exits 2.scripts/next-graph-chain.txt: documents the declaration.planning:wayfinddeclaresskip-to: work-items:decompose, with Q531 (planning: wayfind absorbs the wayfinder #12477) as its reason.scripts/test_check_next_graph.py: one new test proves the declared edge passes and a different stage skip from the same skill still fails. A second proves askip-towith no target exits 2.plugins/session-flow/skills/workflow/context/steps.md: the Decompose stage now says a closed wayfind map whose destination is a set of build items enters it directly (Q531), so the ladder and the gate agree..changes/session-flow/: patch fragment for the ladder text.Verification
Run locally on the merged head:
python3 scripts/check-next-graph.py: clean (it failed with the finding above before the fix)scripts/check-next-graph.test.sh: 19 tests OK, including the 2 new onesscripts/run-ruff.sh checkandformat --checkon both Python files: cleanscripts/check-changelog-fragments.sh --check: all 12 fragments validscripts/check-changed-skills.sh --merge-base: wayfind PASS (0 errors, 6 size and growth warnings, the same ones feat(planning): wayfind absorbs the wayfinder decisions and accept list #12524 reported), workflow PASS (0 errors, 2 warnings)scripts/check-affected.sh --tests: 14 lint checks passed, 1 skipped. Tests: every selected suite passed exceptplugins/planning/skills/setup/scripts/setup-apply.test.sh, which fails the same 96 cases on an unmodified main checkout on this host. That suite setsHOMEto a temp dir, which likely breaks this machine's node resolution (UNVERIFIED), so it is environmental and CI's run decides.scripts/test_check_next_graph.pyruns throughcheck-next-graph.test.sh(above).Not verified: the wayfind eval cases were not run. This lane has no eval runner, as #12524 also noted.
Related
🤖 Generated with Claude Code
https://claude.ai/code/session_01Vcupom3BKF6QuC5U3sRJPj