pm-dispatch SKILL: retire the mandatory shift-end SKILL-suggestion list — narrow the channel to three report classes, default zero (maintainer ruling 2026-08-12) #7950

Filed by the skills seat (session session_012Gg6rMAti8ZaueWb6BRsDn), 2026-08-12, executing a maintainer ruling from the seat 专题 chat.

Provenance (maintainer, verbatim, untranslated)

还有各车道下班时会提交 skills 建议有必要吗?就是那些搞得后来skills越来越乱。

同意

(Approving the seat's recommendation: the mandatory per-shift suggestion list was the growth engine that fattened the skill text — every shift institutionally produced prose candidates. The structural locks against bloat now exist — shrink-only ratchet, ID-lint, ADR-class human merge, single skills seat — so the fix targets the producer side.)

The deltas

1. Retire the fixed deliverable: the seat-exit checklist item "final report includes this tenure's SKILL update-suggestion list (查重后立单)" is removed. A shift with zero suggestions is a good shift; suggestion production is no longer a duty.
2. Narrow the channel to three report classes, default zero — anything a departing seat reports must be one of:
   • ① a principle that is wrong or missing (invariant-level, rare) → to the skills seat via the 专题 channel;
   • ② a mechanizable item (propose a gate or script — ⛔ never prose) → a normal card;
   • ③ a platform-fact change (one row of a references/ fact table).
3. "Lessons learned" prose is no longer admissible into skill texts — under the principles-only regime, error-prevention lives in gates and judgment lives in the model tier; narrative lessons were the old world's compensation mechanism.
4. Reports arrive as finding-class items in the skills lane; this seat triages; the default disposition for anything outside the three classes is close (not planned).

Landing notes

• Edits the seat-exit checklist (in references/seat-post-protocol.md) and the corresponding principle line in SKILL.md. ⚠️ Ratchet headroom 0 — offsets in-PR (this delta is net-negative, which helps).
• Rides the closing consolidated batch (carrier pm-dispatch SKILL Model tiering: mandatory claude-fable-5 for spec SEMANTIC cards — accept/reject behavior or public-surface change (maintainer ruling 2026-08-12) #7945) as its third member — one subagent, one PR.
• Dispatch model: claude-fable-5 (mandatory clause). ADR-class: draft PR, human merge.