luum-home/luum-cognitive-os/coordination-status

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

'Use when you need this Cognitive OS skill: Inspect active multi-session

Install

npx vskill@latest install luum-home/luum-cognitive-os/coordination-status

SKILL.md

Coordination Status

Purpose

Use this skill BEFORE starting work on any file in a session where another Claude/Codex window may be active on the same repo. It answers:

  • Which files are currently held by another session?
  • Who holds the lock, since when, and for what purpose?
  • What does the holder allow (concurrent reads? edits below line N?)?
  • How should I respond if my work conflicts (park / negotiate / escalate)?

This skill is read-only — it inspects the lock registry, never acquires or releases. Use scripts/edit-coop.sh acquire/release for those.

When to invoke

  • Starting a new sub-agent task in a multi-session repo
  • Before editing any file you suspect another agent might be touching
  • When a previous Edit/Write call exited with EDIT-LOCK CONFLICT
  • During post-mortem of a "my changes got reverted" incident

How

bash
bash scripts/cos-coordination-status.sh
# machine-readable:
bash scripts/cos-coordination-status.sh --json | python3 -m json.tool

For file-lock-only detail, keep using:

bash
bash scripts/edit-coop.sh status | python3 -m json.tool

The unified status includes active sessions, task claims, edit locks, stashes, orphan commits, worktrees, pending tasks, and race risks. If a specific linked worktree needs cleanup or porting, switch to /worktree-triage and run scripts/cos-worktree-triage.sh before deleting it. The file-lock-only command returns JSON like:

json
{
"locks": [
{
"target": "tests/conftest.py",
"session": "1777505505-10911",
"agent": "claude-opus-4-7",
"intent": "exclusive-edit",
"since": "2026-04-30T10:15Z",
"heartbeat": "2026-04-30T10:23Z",
"purpose": "ADR-098 quarantine mechanism — Pass 2",
"status": "active"
}
]
}

To check a specific file:

bash
bash scripts/edit-coop.sh check tests/conftest.py

Exit codes: 0 = lockable (free or own), 2 = held by another session.

Decision matrix

Holder stateYour workResponse
Free (no lock)anyacquire and proceed
Own (your session)anyproceed (lock auto-refreshes)
Other session, intent=exclusive-editeditPARK to .cognitive-os/runtime/parked-edits/, do other work
Other session, intent=shared-readreadproceed read-only
Other session, intent=append-onlyappend at endproceed (separate region)
Other session, stale (heartbeat >30min, PID dead)anycall acquire — auto-clears + takes over
Heartbeat live, expires_at >30min awayeditNEGOTIATE via .cognitive-os/runtime/edit-negotiations/
Critical bugfix priorityeditESCALATE: set COS_BYPASS_EDIT_LOCK=1, audit-log proceed

Park protocol

If you decide to PARK:

bash
mkdir -p .cognitive-os/runtime/parked-edits/$YOUR_SESSION
cat > .cognitive-os/runtime/parked-edits/$YOUR_SESSION/conftest.py.json <<EOF
{
"target_file": "tests/conftest.py",
"intended_change_summary": "Add quarantine pass to pytest_collection_modifyitems",
"diff": "...",
"blocked_on": "session=1777505505",
"since": "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
}
EOF

Then move to non-conflicting work. A future hook can drain this queue when the lock releases.

Negotiate protocol

If you need partial access (e.g., they're editing line 200, you need line 50):

bash
mkdir -p .cognitive-os/runtime/edit-negotiations/$THEIR_SESSION
cat > .cognitive-os/runtime/edit-negotiations/$THEIR_SESSION/$YOUR_SESSION.yaml <<EOF
from_session: "$YOUR_SESSION"
target_file: tests/conftest.py
need_lines: [45, 60]
duration_estimate_seconds: 120
purpose: "..."
ask: "yield region 45-60 for 2 minutes"
EOF

The holder reads .cognitive-os/runtime/edit-negotiations/$OWN_SESSION/ on every heartbeat (every 5min by default).

Limits

  • This skill does not automatically apply parked edits. That requires a

separate drain hook (deferred until volume justifies it).

  • Negotiations are advisory — the holder decides. No enforced mediation.
  • COS_BYPASS_EDIT_LOCK=1 works regardless of skill; this skill cannot

prevent it (only audit-log it).

Related

  • ADR-098 — Multi-Agent File Coordination (this skill is its introspection layer)
  • ADR-089 — Multi-Session Git Coordination (sibling layer for git index)
  • scripts/edit-coop.sh — the command-line primitive
  • hooks/edit-lock-pre-tool.sh — PreToolUse[Edit|Write] enforcement
  • templates/edit-conflict-response.md — boilerplate for sub-agents on conflict

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
1
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/luum-home/luum-cognitive-os/coordination-status/badge)](https://verified-skill.com/skills/luum-home/luum-cognitive-os/coordination-status)

Something wrong with this skill or its page? Tell us.