coord-unblocker
Skill0 starsChanged 47 days agoApache-2.0View on GitHub
Install, update, run, or remove the watchdog launchd agent that automatically diagnoses and unblocks coord tasks stuck in needs-brainstorming. Supports /coord-unblocker [--interval N] [--status] [--run-now] [--uninstall].
- Reads credentials
In this repository20 other files it gives its agents.
- coord-archive-indexSkill
- coord-assignSkill
- coord-auditSkill
- coord-checkSkill
- coord-dirtySkill
- coord-discardSkill
- coord-helpSkill
- coord-initSkill
- coord-loop-doctorSkill
- coord-promoteSkill
- coord-refineSkill
- coord-requeueSkill
- coord-reviewSkill
- coord-shape-reviewSkill
- coord-shapeSkill
- coord-splitSkill
- coord-statusSkill
- coord-tokensSkill
- merge-to-mainSkill
- rebase-to-mainSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 4 are distinctive, and 9 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
9
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 16
No args or only --interval N → install/update action.
- Very specific: line 36
Run launchctl list | grep "$LABEL" — output format is PID ExitCode Label.
- Distinctive: line 39
No match → not loaded.
- Very specific: line 40
If $PLIST exists, show StartInterval from it: grep -A1 StartInterval "$PLIST" | tail -1.
- Very specific: line 41
If $LOG exists, show last 30 lines: tail -30 "$LOG".
- Distinctive: line 48
Run one watchdog cycle synchronously:
- Very specific: line 54
Wait for it to finish, then show the last 40 lines of $LOG.
- Very specific: line 74
Use cat > "$PLIST" with a heredoc.
- Distinctive: line 129
Write via a shell heredoc in a Bash call — do not use the Write tool (variables must expand).
- Very specific: line 155
Never touch the main worker plist (com.coord.worker.*).
- Very specific: line 156
Never use sudo — watchdog runs as a user LaunchAgent.
- Very specific: line 157
If coord-project-root fails (not inside a coord project), print the error and stop.
- Distinctive: line 158
Keep output terse — one-line status per step, final summary block.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Manage the per-project watchdog launchd agent (com.coord.watchdog.<project>). The watchdog runs worker/watchdog.sh on a recurring schedule, finds tasks in needs-brainstorming, invokes Claude to diagnose and fix them, and resets fixable tasks to pending.
Argument parsing
Parse $ARGUMENTS (may be empty):
--status— show watchdog state and recent log; stop after reporting.--run-now— execute one watchdog cycle immediately; stop after running.--uninstall— unload and delete the plist; stop after removing.--interval Nor a bare integerN— set interval to N minutes (default: 20).- No args or only
--interval N→ install/update action.
Extract the interval: scan for --interval N pair or a bare integer token. Convert to seconds: INTERVAL_SEC=$((N * 60)). Default N=20 → 1200 seconds.
Shared setup (run before every action)
COORD_TOOLS="${COORD_TOOLS:-$HOME/Projects/coord-wright}"
COORD_MAIN=$("$COORD_TOOLS/bin/coord-project-root")
NAME=$(basename "$COORD_MAIN")
LABEL="com.coord.watchdog.$NAME"
PLIST="$HOME/Library/LaunchAgents/$LABEL.plist"
WATCHDOG="$COORD_TOOLS/worker/watchdog.sh"
LOG="$COORD_MAIN/.coord/watchdog.log"Action: --status
- Run
launchctl list | grep "$LABEL"— output format isPID ExitCode Label.- PID ≠
-→ currently running. - ExitCode ≠
0→ last run failed. - No match → not loaded.
- PID ≠
- If
$PLISTexists, showStartIntervalfrom it:grep -A1 StartInterval "$PLIST" | tail -1. - If
$LOGexists, show last 30 lines:tail -30 "$LOG". - Report: loaded/not-loaded, interval, last-run result, log tail.
Action: --run-now
Run one watchdog cycle synchronously:
"$WATCHDOG" "$COORD_MAIN"Wait for it to finish, then show the last 40 lines of $LOG. Report what was fixed, left, or skipped.
Action: --uninstall
launchctl unload "$PLIST" 2>/dev/null || truerm -f "$PLIST"- Confirm:
launchctl list | grep "$LABEL"→ expect no output. - Report: "watchdog unloaded and plist removed for $NAME".
Action: install / update (default)
- Validate watchdog script exists and is executable.
[[ -x "$WATCHDOG" ]]— if not, runchmod +x "$WATCHDOG".
Write the plist.
Use
cat > "$PLIST"with a heredoc. Substitute$LABEL,$WATCHDOG,$COORD_MAIN,$NAME, and$INTERVAL_SEC:<?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <dict> <key>Label</key> <string>LABEL</string> <key>ProgramArguments</key> <array> <string>WATCHDOG</string> <string>PROJ</string> </array> <key>StartInterval</key> <integer>INTERVAL_SEC</integer> <key>RunAtLoad</key> <false/> <key>WorkingDirectory</key> <string>PROJ</string> <key>StandardOutPath</key> <string>PROJ/.coord/watchdog.log</string> <key>StandardErrorPath</key> <string>PROJ/.coord/watchdog.log</string> <key>EnvironmentVariables</key> <dict> <key>PATH</key> <string>HOMEDIR/.local/bin:/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin</string> <!-- REQUIRED for the watchdog to act: the unattended-autonomy --> <!-- acknowledgement (README "Blast radius"). Include this key ONLY --> <!-- when the operator has COORD_UNSAFE_AUTONOMOUS=1 set in the --> <!-- environment of this install/update run; omit it otherwise and the --> <!-- watchdog exits idle every cycle. On update, REMOVE the key again --> <!-- when the acknowledgement is no longer present — regenerating the --> <!-- plist without it is the revocation path. --> <!-- <key>COORD_UNSAFE_AUTONOMOUS</key> --> <!-- <string>1</string> --> <!-- Optional: pin the watchdog Claude invocation to a specific model. --> <!-- Defaults to ${CLAUDE_MODEL} then claude-sonnet-5 when unset. --> <!-- <key>COORD_WATCHDOG_MODEL</key> --> <!-- <string>claude-haiku-4-5-20251001</string> --> </dict> </dict> </plist>Substitute: LABEL →
$LABEL, WATCHDOG →$WATCHDOG, PROJ →$COORD_MAIN, INTERVAL_SEC → computed seconds, HOMEDIR →$HOME. WhenCOORD_UNSAFE_AUTONOMOUS=1is set in the environment of this run, uncomment/include that key with the exact value1; when it is not set, ensure the key is absent from the plist you write (that is how the operator revokes watchdog autonomy). SetCOORD_WATCHDOG_MODELin this plist to override the modelworker/watchdog.shpasses toclaude --model; the resolution order isCOORD_WATCHDOG_MODEL→CLAUDE_MODEL→claude-sonnet-5. The watchdog reads only this plist's environment — it does not read the project's.coord/config.env.Write via a shell heredoc in a Bash call — do not use the Write tool (variables must expand).
- Reload.
launchctl unload "$PLIST" 2>/dev/null || true launchctl load "$PLIST" - Verify.
A line withlaunchctl list | grep "$LABEL"- 0 $LABEL(dash PID, zero exit) confirms it is loaded and idle. If absent, print the launchctl error and stop. - Report.
watchdog installed: $LABEL interval: N minutes ($INTERVAL_SEC s) plist: $PLIST log: $LOG
Rules
- Never touch the main worker plist (
com.coord.worker.*). - Never use
sudo— watchdog runs as a user LaunchAgent. - If
coord-project-rootfails (not inside a coord project), print the error and stop. - Keep output terse — one-line status per step, final summary block.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

