governance
Toggle PAI governance enforcement on/off. Control chain-of-command routing and session-end requirements.
When & Why to Use This Skill
The Governance Control Skill provides a centralized interface to manage AI agent compliance and operational guardrails. It allows users to dynamically toggle PAI governance, enforce chain-of-command routing, and control session-end requirements, ensuring that AI workflows remain secure and compliant without manual configuration changes.
Use Cases
- Compliance Management: Enforcing strict hierarchical routing and human-in-the-loop approvals for production-level changes or security-sensitive operations.
- Rapid Prototyping: Temporarily disabling governance checks to increase development speed during solo exploration or emergency P0 hotfixes.
- Multi-Agent Orchestration: Controlling the 'chain-of-command' to ensure specialist agents are only spawned through authorized coordinator agents.
- Session Auditing: Managing session-end requirements to ensure all tasks are properly documented and finalized according to organizational standards.
| name | governance |
|---|---|
| description | Toggle PAI governance enforcement on/off. Control chain-of-command routing and session-end requirements. |
| model_tier | haiku |
| can_parallel_with | [] |
| must_serialize_with | [startup, startupO, session-end] |
| preferred_batch_size | 1 |
| max_file_context | 20 |
| compression_level | 2 |
| requires_git_context | false |
| requires_db_context | false |
| - pattern | "production|compliance" |
| reason | "Governance changes for production/compliance require human approval" |
Governance Control Skill
Purpose: Toggle governance enforcement without editing config files Trigger:
/governance [on|off|status]Config:.claude/Governance/config.json
Required Actions
When this skill is invoked, Claude MUST immediately execute the following based on arguments:
Parse Arguments
Check the skill invocation arguments:
- No args or
status→ Show status (default) on→ Enable all governanceoff→ Disable all governancechain on/off→ Toggle chain_of_command_enforcementsession on/off→ Toggle session_end_enforcementbypass on/off→ Toggle bypass_allowed_for_single_file
Action: Status Check (default)
If no arguments or status, read the config and display:
cat .claude/Governance/config.json
Then output this table with actual values:
## Governance Status
| Setting | Status |
|---------|--------|
| governance_enabled | [✅ ON or ❌ OFF] |
| chain_of_command_enforcement | [✅ ON or ❌ OFF] |
| session_end_enforcement | [✅ ON or ❌ OFF] |
| bypass_allowed_for_single_file | [✅ ON or ❌ OFF] |
Action: Toggle All (on or off)
If argument is on, update config:
{
"governance_enabled": true,
"chain_of_command_enforcement": true,
"session_end_enforcement": true,
"bypass_allowed_for_single_file": true,
"notes": "Set governance_enabled to false to disable all governance checks"
}
If argument is off, update config:
{
"governance_enabled": false,
"chain_of_command_enforcement": false,
"session_end_enforcement": false,
"bypass_allowed_for_single_file": true,
"notes": "Set governance_enabled to false to disable all governance checks"
}
After updating, confirm: "Governance [enabled/disabled]. All settings updated."
Action: Toggle Specific Setting
If argument is chain on/off, update only chain_of_command_enforcement.
If argument is session on/off, update only session_end_enforcement.
If argument is bypass on/off, update only bypass_allowed_for_single_file.
After updating, show the full status table.
Quick Reference
| Command | Effect |
|---|---|
/governance |
Show status |
/governance on |
Enable all |
/governance off |
Disable all |
/governance chain off |
Allow direct specialist spawning |
/governance session off |
Skip session-end checks |
/governance bypass off |
Require coordinator for ALL tasks |
When to Disable
Disable governance for:
- Emergency P0 fixes (speed over process)
- Solo exploration sessions
- Quick prototyping
Keep enabled for:
- Production changes
- Multi-agent coordination
- Anything touching compliance/security
Related
.claude/Governance/config.json- Raw config file.claude/Governance/HIERARCHY.md- Chain of command/session-end- Session close-out (respects governance toggle)/startupO- Shows governance status at session start