autonomous-run

Executes approved implementation plans slice by slice under governor limits and human merge gates.

6|Updated Jul 10, 2026
One-click install
npx skills add https://github.com/Mozurok/fhorja.dev --skill autonomous-run-mozurok
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: autonomous-run
Source: https://github.com/Mozurok/fhorja.dev/tree/main/.claude/skills/autonomous-run
Command: npx skills add https://github.com/Mozurok/fhorja.dev --skill autonomous-run-mozurok

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running a large approved implementation plan by hand means repeatedly invoking slice execution, checking safety conditions, and deciding when to stop, which is slow and error-prone across long sessions. This Skill drives an approved, waved IMPLEMENTATION_PLAN through a bounded autonomous delivery track with little supervision while never merging code on its own. ## Core Features & Use Cases - Bounded autonomous execution: Dispatches each plan slice to implement-approved-slice as the single writer, bounded by a runtime governor (token/cost ceiling, max-iteration, wall-clock timeout) and a STOP sentinel file. - Escalation routing: Classifies every slice and halts the wave on boundary changes (schema, contract, migration, security) or any test/eval file touch, surfacing them to the human gate instead of auto-advancing. - Dual preconditions and evidence: Refuses to run without an approved plan and a BOOT readiness verdict, and writes an EVIDENCE_MANIFEST.md indexing every verification artifact produced. - Use Case: After approve-plan and autonomous-readiness pass on a multi-wave plan, run this Skill in one supervised session to produce PROPOSED diffs for each safe slice, then route them to approve-proposed and review-hard for the human merge. ## Quick Start Run the autonomous delivery track on my approved implementation plan in the active task folder and stop at any slice that touches tests or boundaries.

Frequently Asked Questions about autonomous-run

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an approved implementation plan autonomously?▼

Invoke autonomous-run after approve-plan has recorded an approval entry and autonomous-readiness has returned a BOOT verdict for the current plan revision. The controller executes slices via implement-approved-slice, emitting PROPOSED diffs bounded by governor limits and a STOP file.

Can autonomous-run merge or deploy code automatically?▼

No. The controller never commits, merges, deploys, or takes any irreversible step. It produces PROPOSED slice diffs and routes them to approve-proposed and review-hard, where a human always performs the merge.

What happens when a slice touches tests or database schema?▼

The slice classifier returns escalate for boundary changes (schema, contract, migration, security) or any test/eval file touch. The run halts at that slice and surfaces it to the human gate rather than auto-advancing.

Does autonomous-run support resuming a stopped run in a new session?▼

Cross-session durable resume is out of scope in v1. If a run cannot finish in one session, it stops cleanly at a slice boundary and records the resume point in TASK_STATE.md; an opt-in detached background mode runs as one continuous session.

Why does autonomous-run refuse to start my plan?▼

It refuses when the plan lacks an approval log entry for the current revision or when RUN_READINESS.md is absent or records NOT-READY. Both preconditions are independent and must hold before the first slice executes.