القائمة
Working with bosun's daemon system for automated workflows. Covers workflow directories, agent spawning, validators, and file-based chaining.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Bosun configuration — models, sandbox, daemon, Pi settings. Use when changing models, editing config, or understanding how bosun is set up.
Bootstrap a new project using bosun as a foundation — via bun dependency (recommended) or git submodule. Creates a downstream project that inherits bosun's multi-agent infrastructure while adding custom agents, skills, and extensions. Works standalone — fetchable by any pi agent via raw GitHub URL.
Search curated markdown memory like sessions, plans, docs, and skills. Use when recalling prior context or looking for relevant historical/project knowledge.
Analyze Pi session JSONL files using jq patterns. Use when extracting metrics, tool usage, costs, or reviewing session history. Load for session export, summarization, or workflow analysis.
Session-based browser plan review primitives for Bosun. Provides a local-first state model, markdown-aware anchor extraction, and review-session persistence for reviewing existing plan files in dedicated CDP browser windows. Triggers: "plan-review", "review this plan", "review markdown plan".
Session-based browser diff review primitives for Bosun. Provides transport- agnostic v1 schemas plus local state and git snapshot helpers for immutable review rounds. Triggers: "diff-review", "reround", "review round", "snapshot-backed review".
| name | bosun-daemon |
| description | Working with bosun's daemon system for automated workflows. Covers workflow directories, agent spawning, validators, and file-based chaining. |
| license | MIT |
| compatibility | pi |
| metadata | {"category":"workflow","version":"2.0"} |
Background automation engine. Discovers workflows, spawns agents or runs scripts, validates results.
trigger fires (schedule, file watcher, manual)
-> input validator (optional, bun .ts, exit 0/1)
-> spawn agent (pi --print) or run script
-> output validator (optional, bun .ts, exit 0/1)
-> retry with feedback if validation fails
-> done
Key principle: Agent is the brain, scripts just validate. All domain logic lives in agent.md files, not TypeScript handlers.
Workflows are directories scanned from three locations (later overrides earlier):
packages/*/workflows/*/ - packaged defaults.pi/workflows/*/ - repo-level customizationsworkspace/workflows/*/ - user-level overrides (gitignored)Each workflow directory contains:
config.toml - trigger, model, prompt, retry settingsagent.md - agent system prompt (for agent workflows)validate-input.ts / validate-output.ts - optional validatorsConfigured in root config.toml:
[models]
lite = "<your-fast-model>" # e.g. gpt-4.1-mini, claude-haiku-4-5
medium = "<your-balanced-model>" # e.g. gpt-5.3-codex, claude-sonnet-4-6
Workflows reference tiers: model = "lite" resolves to the configured model.
| Package | Workflow | Type | Trigger |
|---|---|---|---|
| pi-chronicles | chronicle-analyzer | agent | hourly |
| pi-chronicles | chronicle-scribe | agent | watches analysis dir |
| pi-session-tools | summarize-session | agent | watches session JSONL |
| pi-session-tools | catchup-sessions | agent | hourly + startup |
| pi-session-tools | fill-handoff | agent | watches handoff dir |
| pi-daemon | backup-workspace | script | daily at 2 AM |
# View daemon status
daemon status
# View daemon logs
daemon logs
# Trigger a workflow manually
daemon trigger <workflow-name>
# Reload workflows (after adding/changing)
daemon reload
# Visualize workflow DAG
just workflow-dag
# Stop daemon
daemon stop
Load the meta-workflow-creator skill for full guidance. Quick version:
mkdir -p .pi/workflows/my-workflow
# Create config.toml, agent.md, optional validators
# Daemon auto-discovers on next heartbeat
meta-workflow-creator - Scaffold new workflowschronicle - Chronicle generation detailscontext-management - Handoff and session management