Lock

---
name: lock
version: 3.0.0
description: Enable autonomous co-pilot mode — agent formulates goal and works until done
triggers:
  - "Enable continuous work mode"
  - "Work autonomously"
  - "Don't stop until done"
  - "Keep working through all tasks"
  - "Work toward this goal"
---

# Lock: Autonomous Co-Pilot Mode

**Purpose**: Enable autonomous co-pilot mode. The agent formulates a goal from the user's natural language, defines what "done" looks like, and works until the goal is achieved or it needs human help.

## Instructions

When the user invokes this command:

### Step 1: Formulate the goal

Read the user's message. From their natural language, formulate:

1. **Goal**: A clear, specific objective statement
2. **Definition of Done**: Concrete criteria for when the goal is achieved (e.g. "tests pass", "PR created", "file exists with X content")

Write both to the goal file as a single document:

```bash
mkdir -p .claude/runtime/locks

Then use the Write tool to create .claude/runtime/locks/.lock_goal with content like:

Goal: [clear objective from user's words]

Definition of Done:
- [concrete criterion 1]
- [concrete criterion 2]
- [concrete criterion 3]

Step 2: Enable lock

python .claude/tools/amplihack/lock_tool.py lock

Step 3: Begin working

Immediately start working toward the goal. Do not ask for confirmation. The LockModeHook will use SessionCopilot to monitor progress and provide guidance on each turn.

How it works

1. The hook fires on every provider:request event
2. SessionCopilot reads the session transcript and reasons about progress
3. If the agent is working — no intervention
4. If the agent is idle — injects specific next-step guidance
5. If the goal is achieved — auto-disables lock mode
6. If stuck — auto-disables and escalates to the user

Disabling

Lock mode auto-disables when:

• Goal is achieved (mark_complete)
• Co-pilot needs human help (escalate)
• User runs /amplihack:unlock

Examples

User says: "fix the auth bug and make sure tests pass" → Agent writes goal: "Fix the authentication bug. Definition of Done: auth tests pass, no regressions in test suite." → Agent enables lock, starts working.

User says: "implement OAuth2 login and create a PR" → Agent writes goal: "Implement OAuth2 login flow. Definition of Done: OAuth2 endpoint works, tests cover happy path and error cases, PR created on GitHub." → Agent enables lock, starts working.

User says: "keep going" → Agent writes goal: "Continue working on the current task until all pending items are complete. Definition of Done: all TODO items resolved, tests pass." → Agent enables lock, continues. ```