god-intervention
Skill181 starsChanged 9 months agoMITView on GitHub
Guides God Committee members through executing interventions. Use for repairs, rollbacks, and emergency actions. Triggers on: intervention, repair, rollback, emergency action.
In this repository16 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- architectSkill
- doc-reviewSkill
- god-consensusSkill
- god-memberSkill
- observabilitySkill
- parallel-exploreSkill
- plan-reviewSkill
- prd-converterSkill
- prdSkill
- researchSkill
- roadmapSkill
- skill-creatorSkill
- vision-builderSkill
- visionSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 453
Before intervention, verify your powers:
≈ “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
God Committee Intervention Skill
Purpose
This skill guides God Committee members through executing interventions - from minor repairs to major system modifications.
Intervention Severity Levels
Level 1: Minor Repair
- Fix corrupted files
- Clear stale locks
- Truncate large logs
- Fix permissions
Consensus Required: No
Level 2: Process Control
- Pause execution
- Resume execution
- Terminate specific processes
Consensus Required: No (unless terminating all)
Level 3: Code Modification
- Fix bugs directly
- Update configurations
- Modify skills
- Add documentation
Consensus Required: Recommended
Level 4: Major Rollback
- Revert multiple commits
- Undo PRD progress
- Reset system state
Consensus Required: Yes
Level 5: System Termination
- Stop all execution
- Archive project state
- Full system shutdown
Consensus Required: Yes
Pre-Intervention Checklist
Before any intervention:
## Pre-Intervention Checklist
- [ ] Issue clearly identified and documented
- [ ] Evidence gathered (logs, metrics, observations)
- [ ] Impact assessment completed
- [ ] Rollback plan prepared
- [ ] Consensus obtained (if required)
- [ ] Execution layer paused (if needed)
- [ ] Backup created (if applicable)Executing Interventions
Level 1: Minor Repairs
# Clear stale locks
./scripts/god/powers.sh repair lock
# Clean up worktrees
./scripts/god/powers.sh repair worktrees
# Fix corrupted JSON files
./scripts/god/powers.sh repair json
# Truncate large logs
./scripts/god/powers.sh repair logs
# Fix script permissions
./scripts/god/powers.sh repair permissions
# Clean git state
./scripts/god/powers.sh repair git
# Run all repairs
./scripts/god/powers.sh repair allLevel 2: Process Control
Pausing Execution
# Pause with reason
./scripts/god/powers.sh pause "Investigation needed: failing tests" YOUR_ID
# Check pause status
./scripts/god/powers.sh pause-statusResuming Execution
# Resume execution
./scripts/god/powers.sh resume YOUR_IDTerminating Processes
# Terminate orchestrator
./scripts/god/powers.sh terminate orchestrator YOUR_ID
# Terminate Aha Loop execution
./scripts/god/powers.sh terminate aha-loop YOUR_ID
# Terminate parallel explorer
./scripts/god/powers.sh terminate explorer YOUR_ID
# Terminate specific PID
./scripts/god/powers.sh terminate pid:12345 YOUR_ID
# Force kill (if terminate fails)
./scripts/god/powers.sh kill orchestrator YOUR_IDLevel 3: Code Modification
Direct File Edits
For simple modifications:
# Append to file
./scripts/god/powers.sh modify "path/to/file" append "content" YOUR_ID
# Prepend to file
./scripts/god/powers.sh modify "path/to/file" prepend "content" YOUR_ID
# Replace file contents
./scripts/god/powers.sh modify "path/to/file" replace "new content" YOUR_IDFor complex edits, use your AI capabilities to directly edit files.
Skill Modifications
# Disable a skill
./scripts/god/powers.sh modify-skill skill-name disable YOUR_ID
# Re-enable a skill
./scripts/god/powers.sh modify-skill skill-name enable YOUR_IDConfiguration Updates
Edit .god/config.json or project configurations directly:
# Example: Update quorum
jq '.council.quorum = 3' .god/config.json > tmp && mv tmp .god/config.jsonLevel 4: Major Rollback
Git Rollback
# Soft reset (keep changes staged)
./scripts/god/powers.sh rollback HEAD~3 soft YOUR_ID
# Mixed reset (keep changes unstaged)
./scripts/god/powers.sh rollback HEAD~3 mixed YOUR_ID
# Hard reset (discard all changes)
./scripts/god/powers.sh rollback HEAD~3 hard YOUR_ID
# Rollback to specific commit
./scripts/god/powers.sh rollback abc123 soft YOUR_IDRestore from Stash
./scripts/god/powers.sh restore-stash YOUR_IDPRD Rollback
To rollback PRD progress:
- Update
project.roadmap.json:jq '.prds |= map(if .id == "prd-xxx" then .status = "pending" else . end)' \ project.roadmap.json > tmp && mv tmp project.roadmap.json - Reset story status in PRD:
jq '.stories |= map(.status = "pending")' \ docs/prd/xxx/prd.json > tmp && mv tmp docs/prd/xxx/prd.json
Level 5: System Termination
⚠️ This requires consensus and should be rare.
# Step 1: Pause everything
./scripts/god/powers.sh pause "System termination initiated" YOUR_ID
# Step 2: Terminate all processes
./scripts/god/powers.sh terminate all YOUR_ID
# Step 3: Stop awakener daemon
./scripts/god/awakener.sh stop
# Step 4: Create final state snapshot
./scripts/god/observer.sh snapshot
./scripts/god/observer.sh report
# Step 5: Archive (optional)
git tag -a "god-committee-termination-$(date +%Y%m%d)" -m "System terminated by God Committee"
# Step 6: Log final entry
./scripts/god/observer.sh event "termination" "System terminated by God Committee"Intervention Patterns
Pattern 1: Investigate and Fix
1. Pause execution
2. Take snapshot
3. Investigate issue
4. Fix the problem
5. Verify fix
6. Resume execution
7. Monitor for recurrencePattern 2: Rollback and Retry
1. Pause execution
2. Take snapshot
3. Identify rollback point
4. Execute rollback
5. Clear any bad state
6. Modify approach if needed
7. Resume executionPattern 3: Emergency Stop and Repair
1. Terminate offending process immediately
2. Take snapshot
3. Assess damage
4. Run repairs
5. Notify other members
6. Plan recovery
7. Execute recovery
8. Resume with monitoringPost-Intervention Protocol
After any intervention:
1. Document the Intervention
## Intervention Report
### Intervention ID
int-[timestamp]
### Type
[repair|process_control|code_modification|rollback|termination]
### Severity
[1-5]
### Triggered By
[observation|alert|proposal|emergency]
### Description
[What was done]
### Reason
[Why it was necessary]
### Steps Taken
1. [Step 1]
2. [Step 2]
3. ...
### Outcome
[Success/Partial/Failed]
### Side Effects
[Any unintended consequences]
### Follow-up Required
[Yes/No - if yes, what]
### Lessons Learned
[What we learned]2. Log the Event
./scripts/god/observer.sh event "intervention" "DESCRIPTION"3. Notify Other Members
./scripts/god/council.sh send YOUR_ID "other,members" "directive" \
"Intervention Completed" "SUMMARY"4. Update Status
# Check intervention history
./scripts/god/powers.sh history interventions 10
# Review repair history
./scripts/god/powers.sh history repairs 105. Monitor for Recurrence
Set up a follow-up observation:
# Trigger immediate observation
./scripts/god/awakener.sh random
# Or use a scheduled check type
./scripts/god/awakener.sh scheduled dailyCommon Intervention Scenarios
Scenario: Failing Tests
# 1. Check current state
./scripts/god/observer.sh check
# 2. Pause if needed
./scripts/god/powers.sh pause "Investigating test failures"
# 3. Analyze failures
cat test-results.json | jq '.failures'
# 4. Decide action:
# - Minor fix: edit directly
# - Major issue: rollback
# - Need discussion: create proposal
# 5. Execute fix/rollback
# 6. Verify
npm test # or equivalent
# 7. Resume
./scripts/god/powers.sh resumeScenario: Stuck Process
# 1. Identify stuck process
ps aux | grep -E "(aha-loop.sh|orchestrator|explorer)"
# 2. Check how long it's been running
# 3. Check for output/progress
# 4. If stuck, terminate
./scripts/god/powers.sh terminate pid:XXXXX
# 5. Clean up any locks
./scripts/god/powers.sh repair lock
# 6. Restart if appropriate
./scripts/aha-loop/orchestrator.sh --continueScenario: Corrupted State
# 1. Stop everything
./scripts/god/powers.sh pause "Corrupted state detected"
./scripts/god/powers.sh terminate all
# 2. Assess damage
./scripts/god/powers.sh repair json # Check for JSON issues
git status # Check git state
# 3. Decide rollback point
git log --oneline -20
# 4. Execute rollback
./scripts/god/powers.sh rollback COMMIT_HASH hard
# 5. Run repairs
./scripts/god/powers.sh repair all
# 6. Verify
./scripts/god/observer.sh check
# 7. Resume
./scripts/god/powers.sh resumeSafety Guidelines
Always
- ✅ Document before acting
- ✅ Take snapshots before major changes
- ✅ Verify after intervention
- ✅ Notify other members of significant actions
- ✅ Have a rollback plan
Never
- ❌ Make major rollbacks without consensus
- ❌ Terminate system without documentation
- ❌ Skip verification after fixes
- ❌ Hide interventions from other members
- ❌ Assume fixes worked without testing
Power Status Check
Before intervention, verify your powers:
./scripts/god/powers.sh statusThis shows:
- Which powers are enabled
- Current pause status
- Running processes
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.

