Agent skill
auxiliary-scripts
Auxiliary script management rules for Ralph agents
Stars
163
Forks
31
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/orchestration/auxiliary-scripts-majiayu000-claude-skill-registr
SKILL.md
Auxiliary Script Management
"Auxiliary scripts in
.claude/session/help with automation and cleanup."
Script Classification
Scripts created in .claude/session/ are automatically classified:
| Classification | Pattern | Retention |
|---|---|---|
| Temporary | *-runner.ps1, pending-messages-*.json, *.exit, restart-flag-*.json, *.tmp |
Auto-deleted after 1 hour |
| Reusable | Documented below | Persist across sessions |
| Unknown | Everything else | Manual cleanup required |
Temporary Scripts
Auto-cleaned after 1 hour:
*-runner.ps1- Agent runner scriptspending-messages-*.json- Message delivery files*.exit- Agent exit status filesrestart-flag-*.json- Restart signal files*.tmp- Temporary files
Do NOT rely on these persisting. They will be automatically deleted.
Creating Reusable Scripts
If you create a helper script that provides value across multiple sessions:
- Document it in the "Reusable Scripts" section of your AGENT.md
- Use descriptive naming:
{agent}-{purpose}.ps1 - Add to
AuxiliaryScripts.Reusableinralph-config.ps1 - Explain its purpose so future sessions understand its value
Reusable Scripts Template
powershell
# Script: {AGENT}-{PURPOSE}.ps1
# Purpose: {Brief description}
# Created: {DATE}
# Used by: {Which agents use this}
param(
[Parameter(Mandatory=$false)]
[string]$SomeParameter
)
# Script logic here
Adding to ralph-config.ps1
powershell
$Script:AuxiliaryScripts.Reusable = @{
"{AGENT}-{PURPOSE}.ps1" = @{
Purpose = "Brief description"
Created = "2026-01-19"
UsedBy = @("pm", "developer")
}
}
Documentation in AGENT.md
Maintain a "Reusable Scripts" section in your AGENT.md:
| Script | Purpose | Usage |
|---|---|---|
| (none yet) |
Update this table when creating new reusable scripts.
Cleanup Guidelines
When Deleting Scripts
- Check if active - ensure no agent is currently using it
- Check dependencies - ensure no other scripts depend on it
- Document removal - add note to progress file if significant
When Keeping Scripts
- Document purpose - add comment header explaining what it does
- Date stamp - include creation date
- Version - if modified, track version changes
Common Script Patterns
Message Delivery Script
powershell
# Used by watchdog to deliver messages to agents
param([string]$TargetAgent, [string]$MessagePath)
$message = Get-Content $MessagePath | ConvertFrom-Json
$inbox = ".claude/session/messages/$TargetAgent/"
Copy-Item $MessagePath "$inbox/$([Guid]::NewGuid()).json"
Health Check Script
powershell
# Used to check if agent is responsive
param([string]$AgentName)
$heartbeatFile = ".claude/session/coordinator-state.json"
$state = Get-Content $heartbeatFile | ConvertFrom-Json
$lastSeen = [DateTime]::Parse($state.agents.$AgentName.lastSeen)
$age = (Get-Date) - $lastSeen
if ($age.TotalSeconds -gt 60) {
Write-Warning "Agent $AgentName unresponsive for $($age.TotalSeconds)s"
return $false
}
return $true
Script Permissions
- All scripts in
.claude/session/should be executable by agents - Never create scripts that modify source code without explicit task assignment
- Scripts should only update session files or state files
Reference
- file-permissions.md — What agents can write to
- context-management.md — Context reset procedures
Didn't find tool you were looking for?