kos-patrol

name	kos-patrol
version	1.0.0
description	Daily sweep: runs kos-lint + staleness detection + knowledge-gap analysis, then produces a MEMORY-format digest ready to be appended to OpenClaw's MEMORY.md by the digest-to-memory skill (Sun 22:00 weekly). Also writes a dashboard to ~/brain/agent/dashboards/knowledge-health-<date>.md.
triggers	["kos patrol","daily patrol","brain daily check"]
tools	["list_pages","get_page","backlinks","query","put_page"]
mutating	true

KOS Patrol Skill

Port of KOS v1 patrol-agent + generate-dashboard.sh logic.

When to run

• Daily 08:00 local (registered via skills/cron-scheduler)
• On demand: gbrain reports run kos-patrol (once wired) or direct bun run skills/kos-jarvis/kos-patrol/run.ts

Phases

Phase 1 — Inventory

• gbrain list --json → total pages, distribution by kind
• Count by confidence (high/medium/low)
• Count pages with status=draft and status=deprecated

Phase 2 — Lint

• Invoke kos-lint skill (6-check)
• Capture counts per severity

Phase 3 — Staleness

For each page with kind in {decision, protocol, project}:

• If updated > 180d ago AND status=active: mark stale
• If page has review_after frontmatter and that date is past: mark overdue

Phase 4 — Gap detection

Heuristic: find frequently-mentioned entities without their own page.

• gbrain extract links --dir ~/brain first (refresh link graph)
• For each entity name appearing ≥3 times across pages but no entity page exists → "missing entity" gap

Phase 5 — Dashboard write

Write to ~/brain/agent/dashboards/knowledge-health-<YYYY-MM-DD>.md:

# Knowledge Health — 2026-04-16

## Inventory
- Total pages: 85
- By kind: source=57, concept=13, project=5, ... (from Phase 1)
- Compilation rate (this week): 63%
- Confidence: high=12, medium=45, low=28

## Issues
- 2 ERROR from kos-lint (see below)
- 7 stale pages (updated >180d, status=active)
- 4 overdue review triggers
- 3 missing entity pages (mentioned ≥3 times)

## Recently compiled (last 7 days)
- [concept:context-engineering] grade=A, 4 impact links
- [source:alignment-faking] grade=B, 2 impact links
- ...

Phase 6 — MEMORY digest emit

Emit a short MEMORY-format block that digest-to-memory skill later appends to ~/.openclaw/workspace/MEMORY.md:

[knowledge-os] 2026-04-16 patrol: 85p / 63% compile-rate / 2 lint-ERROR.
  Recent: context-engineering (A), alignment-faking (B).
  Gaps: 3 entities missing pages. Stale: 7 pages >180d.

Written to ~/brain/agent/digests/patrol-<date>.md (one block per day). The digest-to-memory skill picks up the latest on Sun 22:00 and pushes to OpenClaw MEMORY.

Output contract

• ERROR count > 0 → exit 1 (so autopilot can halt)
• WARN count > 0 → exit 2
• Clean → exit 0
• In all cases: dashboard + digest files written

Delegates

• kos-lint/SKILL.md for the six-check
• confidence-score/SKILL.md for per-page scoring
• digest-to-memory/SKILL.md for MEMORY reflux (Week 3)