s-hiraoku/claude-harnesses · Archived

long-running-orchestrator

Run multi-hour or multi-session work safely by combining the ledger, checkpoints, cost-ceiling guard, and resumable progress notes so context loss does not erase progress.

First seen Jun 13, 2026

Installation

$ npx skills add s-hiraoku/claude-harnesses --skill long-running-orchestrator

Stronger alternatives

This repository is archived — consider an actively maintained alternative.

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from s-hiraoku/claude-harnesses · top by installs.

npx skills add s-hiraoku/claude-harnesses

Browse all from s-hiraoku/claude-harnesses

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Repository health

License LICENSE
Default branch main
Open issues 1
Status Archived

Skill metadata

Parsed from SKILL.md frontmatter.

Declared agents claude-code

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 1,899 B
  • docs SUMMARY.md 204 B

History

  1. First seen on skills.sh
  2. First recorded snapshot · 1 installs

SKILL.md

Long-Running Orchestrator

Use this workflow when a task is expected to span context compactions, multiple Claude Code sessions, or simply too long to hold in a single context window.

Workflow

  1. Initialize the ledger.

- Ensure ledger/current.md exists. - Write a single-paragraph objective, the success criteria, and a checklist plan.

  1. Run the work in small, durable steps.

- Make one focused change at a time. - After each meaningful step, append a dated note to ledger/current.md (or run bash scripts/checkpoint.sh). - Update verification status in ledger/verification.md whenever a test or build run produces a meaningful result.

  1. Guard against runaway.

- The cost-ceiling-guard hook (long-running-pack) tracks cumulative tool calls in ~/.claude-harnesses/cost-ledger.json and forces a stop when the threshold is hit. - When a stop is forced, summarize state in ledger/current.md before doing anything else.

  1. Resume cleanly across sessions.

- The first action of any resume session is to read ledger/current.md and ledger/decisions.md. - State briefly what was last done and what comes next.

  1. Complete only when:

- The objective in the ledger is met. - Verification has run, with results recorded. - A summary of changed files, residual risks, and follow-ups is written.

Constraints

  • Never claim progress that is not reflected in the ledger.
  • Never silently widen scope; if the user asks for a related-but-different change, write a new objective rather than extending the current one.

Final Report

  • objective
  • final ledger pointer
  • verification results
  • residual risks and follow-ups