Use when an approved goal must continue unattended across many steps, milestones, context resets, or sessions.
Scanned 8/31/2026
Install to Claude Code
npx -y skills add lawzava/megapowers --skill autonomous-run --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Autonomous Run?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/lawzava-autonomous-run)More formats (shields.io, HTML) on the badges page.
---
name: autonomous-run
description: Use when an approved goal must continue unattended across many steps, milestones, context resets, or sessions.
---
# Autonomous Run
Prefer native goals and scheduling. Select this skill only for a currently
approved autonomous goal with an existing or newly approved charter. A crash,
compaction, ordinary handoff, or harness switch does not inherit authority.
Keep durable state in one ignored `.megapowers/run/<id>/` directory:
- `charter.md`: objective, done criteria, scope, authority, and cap.
- `checkpoint.md`: milestone, workspace, branch or worktree, HEAD, artifact
identities, completed evidence, delegate ownership and expected return
artifacts, blockers, next safe action, remaining effect authority, and UTC
freshness.
- `journal.jsonl`: append-only observed transitions and their evidence.
Update the checkpoint only at a milestone or real state transition. Journal
what an oracle proved, not intent or a progress guess. Derive status from the
journal, checkpoint, and fresh evidence; chat history is not a status oracle.
On resume, reread repository instructions, charter, checkpoint, journal tail,
and current workspace state. Compare the repository, worktree, branch, HEAD,
runtime, and relevant external state. If evidence is missing, contradictory, or
shows a workspace mismatch, stop before acting. Do not execute the next command
until scope and authority are re-established. A native goal does not transfer
automatically across harnesses.
Use `paused` when a cap or intentional stop ends authorized execution. Preserve
the checkpoint and wait for renewed authority. Use `blocked` only for a concrete
external dependency outside current authority, such as a provider limit or
expired credential, with its evidence and unblocking event; surface the block
once rather than waiting silently. Mark done only after every criterion passes its stated oracle. Use
`safe-effects` for every external mutation.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!