Skip to content
sdlc

B-PUHO-s-0006-per-skill-equipping-batches-companion-doc-drift-sweep

Status: closed/abandoned

The per-skill equipping tail, one PR per skill but sharing patterns (define the B-DHKB verification-step convention first).

Description-trigger refresh (rule 4):

  • project-check — stale desc (now 8 checks, not 1)
  • setup — add sdlc.yaml/.gitignore/flags + “update path” context
  • update-skill-doc — add drift-hook WHEN
  • find-quality-checks — add configure-quality-gates WHEN
  • migrate-runtime-state — add post-upgrade WHEN

Verification-step + examples (rules 7+10): docs, backlog-triage, principle-review, find-quality-checks, dev-update-docs, milestones-from-file.

Low residue / progressive-disclosure trims (NOT standalone):

  • task-close-out — recast non-safety negatives; keep pinned safety ones
  • task-define — extract the 94-line gap-report schema narration; front-load the main-branch guardrail

Companion-doc drift sweep via the update-skill-doc op:

  • entities-audit — doc says --type vs canonical --types
  • find-quality-checks — generated doc still has “does NOT stage” negatives
  • task-close-out — invariants.yaml prose "DONE pr=" vs the namespaced marker

NOTE: rule-11 eval gaps are DEFERRED to existing B-5PY9 (“evals for skills”) / T-RXKE (“eval harness for doc-authoring skills”) — build no per-skill evals; at most a one-line Notes pointer.

Closed 2026-08-16: superseded by D-VSLI-distributed-work-runner-architecture. The skill-prose lint retires at cutover with the lifecycle skills, and SKILL.md becomes a generated head over a process definition, so S-0006’s sequencing rules go inert.

← Back to Backlog