agentleFS
zjp1997720zjp1997720/zhijian-skills/skills/codex-handoff/SKILL.md

codex-handoff

Skill787 starsChanged 4 months agoMITView on GitHub

Roll an oversized or slow Codex task into a fresh user-visible task with compact continuity. Use only for an explicit handoff or context rollover; not for ordinary continuation, delegation, full-history forks, Git migration, or handoff files.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 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.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Write a self-contained prompt of roughly 2,000–4,000 Chinese characters or comparable size.

    line 17
  2. Very specific:

    Call create_thread with the compact prompt.

    line 32
  3. Distinctive:

    Stop work in the old task.

    line 35
  4. Very specific:

    Do not use fork_thread; full-history copies defeat the context reset.

    line 39
  5. Distinctive:

    Do not spawn a subagent; delegation is not a user-owned continuation task.

    line 40
  6. Very specific:

    Do not use handoff_thread.

    line 41
  7. Distinctive:

    Do not create or resume a handoff document.

    line 42
  8. Distinctive:

    Do not trigger on ordinary requests to continue in the current task.

    line 43

≈ “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

Codex Handoff

Preconditions

  • Explicit invocation approves one user-owned task. Arguments set its focus; otherwise recover the current objective from verified context.
  • Require list_projects and create_thread. If unavailable, stop; do not substitute a shell command or Markdown file.

Build the continuity prompt

Write a self-contained prompt of roughly 2,000–4,000 Chinese characters or comparable size. Prefer smaller when links carry detail. Include:

  • objective and first action;
  • completed work and current state;
  • decisions, constraints, and material rejected alternatives;
  • authoritative files, plans, issues, commits, diffs, or URLs;
  • verification performed and failures observed;
  • remaining work, blockers, fragile state, and transient paths;
  • relevant installed Skills.

Point to authoritative artifacts instead of reproducing them. Omit transcript history, large diffs or status output, repeated reasoning, secrets, credentials, and unrelated personal information. Mark transient paths as fragile. End with: read active instructions, verify current state, and begin the first action now.

Create the fresh task

  1. Resolve the current saved project with list_projects; obey project and host environment rules. Never guess another project or silently fall back to projectless. If the target omits material uncommitted state, stop and state the risk.
  2. Call create_thread with the compact prompt. Omit model and reasoning overrides unless explicitly requested.
  3. For a ready threadId, title it with set_thread_title and take one bounded wait_threads snapshot; do not wait for completion. For only clientThreadId, report setup as queued, not running.
  4. Return a one-line summary and the created-task directive. Claim only the returned state.
  5. Stop work in the old task. Keep it as history; do not archive it or keep editing the shared workspace.

Exclusions

  • Do not use fork_thread; full-history copies defeat the context reset.
  • Do not spawn a subagent; delegation is not a user-owned continuation task.
  • Do not use handoff_thread; it moves another task's Git environment.
  • Do not create or resume a handoff document; that is a file-oriented workflow.
  • Do not trigger on ordinary requests to continue in the current task.

Maintain the manual trigger boundary with evals/trigger_cases.json and the continuity behavior with evals/output/cases.jsonl.

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.