Skip to content
Back to skills

Daemon Init

ASecurity

Initialize AIWG daemon mode configuration from a named profile for persistent background sessions

  • 211 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 21, 2026
ai-agentsbashexpressdocumentation

Works with

  • cli

Security analysis

A100/100

Scanned September 21, 2026

npx -y skills add jmagly/ai-writing-guide --skill daemon-init --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Daemon Init?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Daemon Init
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/jmagly-daemon-init-39373cf7/badge)](https://www.skillsdirectory.com/skills/jmagly-daemon-init-39373cf7)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
namespace: aiwg
name: daemon-init
platforms: [all]
description: Initialize AIWG daemon mode configuration from a named profile for persistent background sessions

---

# Daemon Init

You initialize AIWG daemon mode configuration from a named profile. The daemon is a persistent AI session that handles background task execution, messaging integration, and multi-loop orchestration — running continuously rather than responding to individual prompts.

## Triggers

Alternate expressions and non-obvious activations (primary phrases are matched automatically from the skill description):

- "start daemon mode" → init with default profile
- "set up background agent" → init with manager profile
- "configure aiwg for continuous operation" → init with orchestrator profile
- "i want aiwg running in the background" → init with manager profile
- "reinitialize daemon" → init with `--force`

## Trigger Patterns Reference

| Pattern | Example | Action |
|---------|---------|--------|
| Default init | "initialize daemon" | Run `aiwg daemon-init` |
| Named profile | "init daemon as orchestrator" | Run `aiwg daemon-init orchestrator` |
| Force reinit | "reinitialize daemon config" | Run `aiwg daemon-init --force` |
| Worker mode | "set up a worker daemon" | Run `aiwg daemon-init worker` |
| Monitor mode | "run daemon in monitoring mode" | Run `aiwg daemon-init monitor` |

## Behavior

When triggered:

1. **Extract intent**:
   - Is a profile name specified? Default to `manager` if not.
   - Is `--force` needed (user mentioned reinitializing or overwriting)?

2. **Select the appropriate profile**:

   | Profile | Authority | Use Case |
   |---------|-----------|----------|
   | `manager` (default) | Balanced — orchestrate and execute | General-purpose persistent agent |
   | `orchestrator` | High — coordinate other agents | Driving multi-agent workflows |
   | `worker` | Low — execute, minimal decisions | Background task execution |
   | `monitor` | Read-only — observe and alert | Ops monitoring, health checks |

3. **Run the command**:

   ```bash
   # Default profile (manager)
   aiwg daemon-init

   # Named profile
   aiwg daemon-init orchestrator
   aiwg daemon-init worker
   aiwg daemon-init monitor

   # Overwrite existing config
   aiwg daemon-init --force
   aiwg daemon-init orchestrator --force
   ```

4. **Know where config is written**: `.aiwg/daemon/config.yaml` in the current project. This file controls:
   - Session persistence behavior
   - Messaging integration endpoints (Telegram, Slack, etc.)
   - Task queue and loop settings
   - Authority level and decision boundaries

5. **Report the result** — Confirm which profile was initialized and the config path. If daemon is already initialized, remind the user to use `--force` to overwrite.

## Examples

### Example 1: Default initialization

**User**: "Set up daemon mode"

**Extraction**: Init request, no profile specified — use `manager`

**Action**:
```bash
aiwg daemon-init
```

**Response**: "Daemon initialized with `manager` profile. Config written to `.aiwg/daemon/config.yaml`. To start the daemon, see `docs/daemon-guide.md` for session launch instructions."

---

### Example 2: Orchestrator profile

**User**: "Initialize the daemon as an orchestrator — I want it coordinating multiple sub-agents"

**Extraction**: Init with `orchestrator` profile

**Action**:
```bash
aiwg daemon-init orchestrator
```

**Response**: "Daemon initialized with `orchestrator` profile. High-authority orchestration settings applied. Config at `.aiwg/daemon/config.yaml`. The orchestrator profile enables `aiwg mc` (Mission Control) dispatch patterns."

---

### Example 3: Force reinitialize

**User**: "Reinitialize the daemon — the old config is stale"

**Extraction**: Force reinit, no profile specified — use `manager` with `--force`

**Action**:
```bash
aiwg daemon-init --force
```

**Response**: "Daemon config overwritten with `manager` profile. Previous config backed up to `.aiwg/daemon/config.yaml.bak`."

---

### Example 4: Worker for CI pipeline

**User**: "I need a worker daemon for running background test loops in CI"

**Extraction**: Init with `worker` profile

**Action**:
```bash
aiwg daemon-init worker
```

**Response**: "Daemon initialized with `worker` profile. Minimal decision authority configured — this profile executes tasks and reports results without autonomous orchestration. Suitable for CI background loops."

## Clarification Prompts

If the user's intent is ambiguous:

- "Which daemon profile should I use? Options: `manager` (default), `orchestrator`, `worker`, `monitor`"
- "A daemon config already exists at `.aiwg/daemon/config.yaml`. Should I overwrite it with `--force`?"
- "Are you initializing for a new project, or reconfiguring an existing daemon?"

## References

- @$AIWG_ROOT/src/cli/handlers/daemon.ts — Daemon command handler
- @$AIWG_ROOT/docs/daemon-guide.md — Full daemon documentation including session launch
- @$AIWG_ROOT/docs/messaging-guide.md — Messaging integration (Telegram, Slack) configuration
- @$AIWG_ROOT/docs/cli-reference.md — CLI reference

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.