skills-scheduled-mirror-sync — 2026-05-30

What happened

Daily skills-scheduled-mirror-sync. Compared canonical Projects/Skills/<X>/SKILL.md against the Claude/Scheduled/ paths that scheduled tasks load.

Mirror map results

Mirror	Mode	Result
`_post-run-qa-reflection.md`	partial	CLEAN
`sail-kb-daily-pulse/SKILL.md`	full-skill	DRIFT — synced (preload headers identical, canonical body restored to dest)
`sail-kb-daily-pulse/build_report.py`	binary	CLEAN

Drift detail. The destination's working tree had a stripped frontmatter version of sail-kb-daily-pulse/SKILL.md (lost brittleness, last_successful_run, deliverable_path, data_source_ladder). HEAD at commit 9e2a109 already had the rich frontmatter (matched canonical at last sync 2026-05-26). After this run, working tree matches HEAD again — so no commit was needed.

Git working-tree heads-up (Claude/Scheduled)

13 dirty entries in ~/Documents/Claude/Scheduled total. Only sail-kb-daily-pulse/SKILL.md was on the mirror map (now clean). The other 12 are unrelated and were left alone per the explicit-add rule. Worth a cleanup pass by the owning agent.

Skills_Library compatibility surface

139 Skills, 1 Skills_Library entry. 0 drift in existing mirrors. 123 unmirrored — not auto-created per 2026-04-24 deprecation (commit 79035d2).

Cross-tool skill surface check

No blocking flags. Two informational items:

skill_surface_check.py detects Projects root as …/routines-build (subfolder), so the Projects/Skills count reads 11 instead of actual 139. Output flags are still accurate but counts are misleading.
AGENT-TOOL-MANIFEST.md reported missing — path may have moved.

QA: Open

skill_surface_check.py Projects-root detection is buggy. Recommended fix: walk up from the script's location looking for a top-level marker (FOLDER-INDEX.md, AGENTS.md, CLAUDE.md), or accept a --projects-root flag.
AGENT-TOOL-MANIFEST.md reference path needs updating (or the file needs to exist at the expected location).
12 unrelated dirty files in Claude/Scheduled working tree — accumulating across recent runs. Suggest the owning agent (daily-briefing, daily-repo-health-check, etc.) sweep them.

Verification commands run

shasum -a 256 <canonical> <dest>      # before sync
python3 /tmp/sail-mirror-sync2.py     # dry-run
python3 /tmp/sail-mirror-sync2.py --apply
git status --porcelain -- _post-run-qa-reflection.md sail-kb-daily-pulse/SKILL.md sail-kb-daily-pulse/build_report.py

Post-sync dest sha for sail-kb-daily-pulse/SKILL.md: 6eb89e8d7f31ab95cca57dc22eda5c8a08965dbd42164a2d719770beaafa4f5a (matches canonical exactly).