mission

name	mission
description	Autonomous mission execution system. Decomposes complex goals into milestones with DAG-scheduled tasks, role-based agents (planner/coder/reviewer/tester), and milestone validation. Use when the user describes a complex multi-step task that needs structured decomposition.

Mission System (Autonomous Goal Execution)

You are a Mission Controller. You can decompose complex goals into structured plans with milestones, tasks, and role-based agents, then execute them autonomously with DAG scheduling.

When to use missions vs tasks:

Scenario	Use
Single task, one agent	orchestrator skill
Quick parallel tasks, flat structure	orchestrator skill (with plan)
Complex goal needing decomposition	mission skill
Multi-phase work with milestones	mission skill
Need code review + testing built in	mission skill

Context Management Rules

Missions can produce extensive output. To avoid context blowup:

1. Use status — get compact progress summary (milestones + tasks + timing)
2. Use list — get 1-line-per-mission overview
3. Use log --tail N — only last N log entries
4. Never dump full logs — always use --tail
5. Don't poll too frequently — missions take minutes/hours, check every 2-5 minutes

---

Tool: Mission CLI Script

All operations use the same script with subcommands:

python3 prompts/skills/mission/scripts/mission.py <command> [options]

---

Commands Reference

plan — Decompose Goal into Structured Plan

Generate a mission plan without executing. The user reviews the plan before approval.

python3 prompts/skills/mission/scripts/mission.py plan "Implement JWT auth with refresh tokens" \
  --workspace ~/Projects/api

Output shows milestones, tasks, roles, and dependencies. Status will be planned.

start — Plan + Auto-Approve + Execute

Skip the review step — plan and immediately begin execution:

python3 prompts/skills/mission/scripts/mission.py start "Add comprehensive test suite" \
  --workspace ~/Projects/myapp

approve — Approve a Planned Mission

After reviewing a plan, approve it to start execution:

python3 prompts/skills/mission/scripts/mission.py approve msn-a1b2c3d4

status — Mission Progress

Get detailed progress with milestones, tasks, timing, and cost:

python3 prompts/skills/mission/scripts/mission.py status msn-a1b2c3d4

list — List All Missions

Compact overview of all missions:

python3 prompts/skills/mission/scripts/mission.py list
python3 prompts/skills/mission/scripts/mission.py list --include-completed

cancel — Cancel a Mission

python3 prompts/skills/mission/scripts/mission.py cancel msn-a1b2c3d4

pause — Pause a Running Mission

python3 prompts/skills/mission/scripts/mission.py pause msn-a1b2c3d4

resume — Resume a Paused Mission

python3 prompts/skills/mission/scripts/mission.py resume msn-a1b2c3d4

log — View Mission Log

# Last 10 entries (recommended)
python3 prompts/skills/mission/scripts/mission.py log msn-a1b2c3d4 --tail 10

# All entries (use sparingly)
python3 prompts/skills/mission/scripts/mission.py log msn-a1b2c3d4

---

Workflow Pattern

Standard: Plan -> Review -> Approve -> Monitor -> Report

# 1. Decompose the goal
python3 .../mission.py plan "Implement user authentication with OAuth2" --workspace ~/Projects/api

# 2. Show the plan to the user, explain milestones and tasks
#    (The plan output includes milestone/task breakdown)

# 3. User approves -> start execution
python3 .../mission.py approve msn-XXXXXX

# 4. Monitor progress (every 2-5 minutes)
python3 .../mission.py status msn-XXXXXX

# 5. When complete, report the summary to the user

Quick Start (trusted goals)

# Plan + execute in one step
python3 .../mission.py start "Refactor database layer to use connection pooling" --workspace ~/Projects/api

# Monitor
python3 .../mission.py status msn-XXXXXX

---

Mission Structure

A mission consists of:

• Milestones — logical phases (e.g., "Setup", "Implementation", "Testing")
  • Milestones can depend on other milestones (DAG scheduling)
  • Each milestone has validation criteria and commands
• Tasks — individual work items within milestones
  • Each task has a role: planner, coder, reviewer, tester
  • Tasks can depend on other tasks within the same milestone
  • Tasks run in parallel when dependencies allow
• Roles — specialized agent behavior:
  • planner: Analyzes requirements, designs architecture
  • coder: Implements code changes
  • reviewer: Reviews code for quality and correctness
  • tester: Writes and runs tests

---

Rules

1. Always plan first — use plan to generate the decomposition before executing
2. Show plan to user — always present the plan and ask for confirmation before approve
3. Don't poll too frequently — missions take minutes to hours; check every 2-5 minutes
4. Use --tail on logs — never dump full mission logs into context
5. Report milestones — when a milestone completes, report it to the user
6. Handle failures — if a mission fails, check the status and log for details before retrying

---

Error Handling

Error	Cause	Action
[ERROR] Mission not found	Invalid mission ID	Check with list
[ERROR] Import failed	nexus not installed	Install with pip install -e ../nexus
[ERROR] Mission not in planned status	Already running/completed	Check status first
[ERROR] API key not set	Missing OPENAI_API_KEY	Set env var

---

Environment Variables

Variable	Default	Description
OPENAI_API_KEY	(required)	LLM API key
OPENAI_API_BASE	(OpenAI default)	API base URL
NEXUS_MODEL	gpt-4o	Default model for task execution
NEXUS_WORKSPACE	~/Projects	Default workspace directory