# Auxiliary Scripts

> Auxiliary script management rules for Ralph agents

- Skill: `diegosouzapw/auxiliary-scripts` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add diegosouzapw/auxiliary-scripts`
- Raw SKILL.md: https://api.skillmd.com/api/skills/diegosouzapw/auxiliary-scripts/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: diegosouzapw (https://skillmd.com/u/diegosouzapw)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/diegosouzapw/auxiliary-scripts

---


# 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 scripts
- `pending-messages-*.json` - Message delivery files
- `*.exit` - Agent exit status files
- `restart-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:

1. **Document it** in the "Reusable Scripts" section of your AGENT.md
2. **Use descriptive naming**: `{agent}-{purpose}.ps1`
3. **Add to `AuxiliaryScripts.Reusable`** in `ralph-config.ps1`
4. **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

1. **Check if active** - ensure no agent is currently using it
2. **Check dependencies** - ensure no other scripts depend on it
3. **Document removal** - add note to progress file if significant

### When Keeping Scripts

1. **Document purpose** - add comment header explaining what it does
2. **Date stamp** - include creation date
3. **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](file-permissions.md) — What agents can write to
- [context-management.md](context-management.md) — Context reset procedures