agentleFS
anthony-chaudharyanthony-chaudhary/fak/.agents/skills/run-it-all-night/SKILL.md

run-it-all-night

Skill49 starsChanged 14 days agoApache-2.0View on GitHub

Plan and run one bounded overnight issue worker through fak's guarded dispatch path, with typed capacity admission, lane leases, explicit dry-run/live gates, and independent git/DOS/test reconciliation. Use when the operator asks to let a narrow repo task run unattended overnight.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run one narrow, issue-bound task overnight without bypassing the kernel's account, lease, or witness gates.

    line 11
  2. Distinctive:

    Start from the current repository root.

    line 16
  3. Distinctive:

    Run the authoritative dispatch preflight and honor every typed refusal, account park, retry-after, session ceiling, and seat-depletion …

    line 27
  4. Distinctive:

    If another live holder owns the tree, stop this unit and choose non-overlapping work.

    line 31
  5. Distinctive:

    If the worker exits without a commit, posts a typed refusal, or makes no progress, record the failure receipt on the issue.

    line 84

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Run one narrow, issue-bound task overnight without bypassing the kernel's account, lease, or witness gates.

1. Resolve the workspace and issue

Start from the current repository root; never hard-code a checkout path. Confirm that the task has one dedicated issue, a bounded done condition, likely files, and a focused acceptance witness. If any of those are missing, repair the issue before launch.

git rev-parse --show-toplevel
gh issue view <issue> --json number,title,state,body,url

2. Capture capacity and ownership evidence

Run the authoritative dispatch preflight and honor every typed refusal, account park, retry-after, session ceiling, and seat-depletion result. Never bypass a refusal with Claude dangerous-mode flags or an unguarded process launch.

Acquire the issue's exact file tree through dos lease-lane before any edit or live dispatch. If another live holder owns the tree, stop this unit and choose non-overlapping work.

3. Dry-run before live launch

Render the issue-specific guarded command first. A dry-run must show the intended worker identity, bounded fuel, file tree, account admission, and zero launches.

For general backlog issues, plan the bounded unit via fak issue-queue or fak dispatch tick:

fak issue-queue --lane <lane> --top 1 --json
fak dispatch tick --issue <issue_id> --workspace . --json

For stale-work units, use the issue-bound loop and keep issue creation separate from process launch:

fak stale-work loop --packet <packet.json> --issues <issues.json> --max-wave 1 --json

Inspect the plan. Do not add --live-launch or --live until the issue is contract-valid, the typed capacity preflight admits it, and the lane lease is held.

4. Launch one bounded worker

Enable exactly one live gate for exactly one worker. Prefer the issue-specific command rendered by the dry-run:

fak dispatch tick --issue <issue_id> --workspace . --live --json

For stale-work units:

fak stale-work loop --packet <packet.json> --issues <issues.json> \
  --live-launch --max-wave 1 --json

A wider wave is allowed only after collision pricing proves disjoint trees and the account/seat preflight admits the full width. Overnight duration is not permission for unbounded turns, workers, retries, or spend.

5. Monitor and reconcile from independent witnesses

Monitor the dispatch receipt and run archive, but never accept the worker's narration as proof. A ship requires all of:

  • the expected scoped git diff and origin ancestry;
  • the issue-specific test or read-back green;
  • dos commit-audit returning OK / diff-witnessed; and
  • independent issue state agreeing with the witnessed commit.

If the worker exits without a commit, posts a typed refusal, or makes no progress, record the failure receipt on the issue. Do not relaunch until the cause is repaired and preflight is green.

6. Finish safely

Push promptly once the exact issue is green. Release the lane lease only after publication or a recorded non-ship. Before launching another worker, repeat capacity preflight, dry-run, and the full witness barrier.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.