Repository navigation
planning: interview restate, wayfind absorb, stress test per mode #7586
Description
Parent
Refs #7555. Source: PLAN Phase 18, topic poteto-review. Thread: planning.
Depends on
- config-cascade: settings resolution and convention home #7571 Settings resolution and convention home
- skill-quality: skill authoring, quality and catalog #7582 Skill authoring, quality and catalog (final plugin shape before the devils-advocate eval)
Outcome
/planning:interview gains a restate action that waits for a yes and closes Matt's grilling
gaps; /planning:wayfind absorbs Matt's wayfinder; the stress-test setting is keyed per mode and
settled by an eval.
Decisions
/planning:interview: a "because" on the goal, a run-it-to-settle-it step, the quality bar when
the user can't judge; warn against nodding along; cross-check stated facts against code; route
"document my repo" to curate-language (D5, D69, Q72). Origin: Matt's grilling.- Interview
restateaction: one paragraph (goal, problem, done), waits for a yes in interactive
sessions (D104, Q130, Q131, Q170; fixed floor; Matt's). - feat(planning): every interview question opens with the context that drives the decision, with links to verify it #7475 and feat(planning): interview pairs every code with its plain name at every use #7480; one-question-at-a-time as a user setting; probe topics as prose in the planning
convention; offer/planning:audit-answersafter two accept-all rounds (setting, 0 = never);
end-of-round index table dropped (D69, D78, Q70, Q71, Q73, Q517). /planning:wayfindabsorbs Matt's wayfinder and reports every conflict: interview is the entry;
chart runs the interview inline; map closes to prd, decompose, work; single-session efforts skip the
map; destination confirmed first; setting research auto-startauto(D5, D108, Q205, Q531, Q532).
Wayfind accept items per Q523. Setting: wayfind HITL concurrency 1; suggestion threshold to be
measured./planning:devils-advocatestays in planning;plan_stress_testkeyed per mode (plan, proposal,
incumbent) (D103, Q164).plan-reviewerstays on, configurable (Q166; ours).- Type rules move from
/planning:designto a type-rules convention synced into plugins (Q12). - Research tags: devils-advocate eval (12 plans, 3 arms, 3 runs) after the plugin shape is final
settles theplan_stress_testdefault (Q165); hand-off phrasing eval (Q188); eval variance checked
on the first run (assumption 10).
Key interfaces
From the design's module layout: "planning: /planning:interview gains restate;
/planning:wayfind absorbs Matt's wayfinder; /planning:devils-advocate stays; plan-reviewer
stays on. Type rules leave /planning:design for a type-rules convention synced into plugins."
Settings: "Offer answer audit after 2 accept-all rounds (0 = never). ... plan_stress_test per mode;
plan-reviewer on." "Research auto-start auto; wayfind HITL concurrency 1."
Slices, filed when this phase starts: interview-and-restate, wayfind-absorb,
stress-test-per-mode-and-eval, type-rules-convention.
Done when
-
bash plugins/skill-quality/scripts/check-skill.sh plugins/planning/skillsexits 0, and the
interview skill has arestateaction file (noactions/directory on main). - The devils-advocate eval slice carries a
Probe result:comment before the default changes. - Every slice sub-issue is closed by a merged PR.
Out of scope
- Decompose and work-items changes (work-items thread).
Related
- fix(planning): interview page wrap-up friction: stale closed question, id count confusion, ops re-apply, restate cap #5893 interview wrap-up friction (sub-issue of this thread).
- feat(planning): interview questions must be self-contained, with an agreed terminology matrix and references #6136 self-contained interview questions (sub-issue of this thread).
- feat(planning): every interview question opens with the context that drives the decision, with links to verify it #7475 interview questions open with their context (sub-issue of this thread).
- feat(planning): interview pairs every code with its plain name at every use #7480 interview pairs every code with its plain name (sub-issue of this thread).
Activity
github-actions commentedon Oct 16, 2026
Proposed work class: work-class: structural
Basis: automated intake triage; the lanes App applied this class.
Size: Large
Complexity: Medium
Uncertainty: Knowable
Flags: none
Blocked by: none
#7586 (comment)
This decision settles the open questions. The work is too big for one change, so the first step is filing the four slices. The paid eval draws on prepaid credit, and the host run is routed separately, so neither holds the issue.
Automated triage, applied by the lanes App.
28 remaining items
kyle-sexton commentedon Nov 2, 2026
Decision recorded.
All of this issue's blockers are closed: #7571, #7938, and the #7582 slices #11955, #11965 and #11971. The edges moved to those slices in the #7582 split comment. None of its own done-when is met on origin/main:
plugins/planning/skills/interview/has noactions/directory and norestateaction. Four of its five sub-issues closed completed: #5893, #6136, #7475, #8647. #7480 closed as not planned, and this draft does not count it as delivered. A lane run stopped on the issue withsplit-requiredon 2026-11-02 (run 41374965540), and the four slices the body names were never filed. The first step is to file them, under owner decision Q-F. The devils-advocate eval (12 plans, 3 arms, 3 runs) is a paid eval run that settles theplan_stress_testdefault, so this issue needs a host run.Acceptance criteria:
bash plugins/skill-quality/scripts/check-skill.sh plugins/planning/skillsexits 0, and the interview skill has arestateaction file.Probe result:comment before the default changes.Route: host-run
Posted by the issue-triage session. This was generated by AI.
kyle-sexton commentedon Nov 3, 2026
Finding from the #5769 host run (interview evals): one case scores lower with the plugin than without it.
final-confirm-active-replyscored 0.56 with the planning plugin and 0.89–1.00 without it. That was 3 runs per arm, opus judge, subscription login, CLI 2.1.316. The gap was the same before #12321 (parent4b042c64b8) and after it (bacba0349c), so #12321 did not cause it; it predates that change. The full run record is on #5769: #5769 (comment)Recorded here because this issue covers the interview skill's restate and confirm behaviour. Next step: read the case's with-plugin traces to find which instruction makes the final confirmation worse, then fix it or narrow the case.
kyle-sexton commentedon Nov 4, 2026
Closing: every done-when item holds on origin/main (checked 2026-11-04).
bash plugins/skill-quality/scripts/check-skill.sh plugins/planning/skillsexits 0 on a detached origin/main worktree (12 passed, 0 failed).plugins/planning/skills/interview/actions/restate.mdexists.Probe result:comments (1, 2), and feat(planning): add plan_stress_test per mode with its 12-plan eval suite #12835 merged it.The open finding from the #5769 run is outside these criteria. It moves to #13003.
Posted by host-run Lane 5. This was generated by AI.