agentleFS
glebisglebis/claude-skills/skills/handoff/SKILL.md

handoff

Skill410 starsChanged 5 months agoMITView on GitHub

Hand a written brief to an ACP agent that works in its own git worktree and herdr workspace, then check its work before anyone merges it. Writes a dated handoff doc, creates a beads epic when the repo has one, cuts a `handoff/<id>-<slug>` branch from the current branch, launches the agent through herdr''s `acp-run` (default `cmd acp`), and later verifies commits, stray changes and the brief''s checks. Use when the user says "hand this off", "create a handoff", "start it in its own worktree", "give this to an agent", "launch an agent on this epic", or wants an epic or multi-step to-do worked on away from a shared checkout. Requires git and a running herdr; beads is optional.

191 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use decision-mapping, setup-ts-deep-modules and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • FabienDehopreSkill

    decision-mapping

    Turn a loose idea into a sequenced map of investigation tickets, then drive them to resolution one at a time.

    Also in 8 projects that use this

  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 42 projects that use this

  • AWSSkill

    to-questionnaire

    Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

    Also in 93 projects that use this

  • GoogleSkill

    ask-matt

    Ask which skill or flow fits your situation. A router over the skills in this repo.

    Also in 87 projects that use this

  • RoentekSkill

    loop-me

    Grill me about specs for the workflows I want to build, within this workspace.

    Also in 51 projects that use this

  • LesaniSkill

    writing-beats

    Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.

    Also in 54 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 172 edited versions we read keep these.

  • 1

    Distinctive

    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. Very specific:

    Run them from inside the target repository.

    line 26
  2. Very specific:

    Write the doc: Use handoff new "<title>" or the project's own convention: look at existing handoff or spec docs first and match their …

    line 31
  3. Very specific:

    Close: after the merge: handoff close ID removes the worktree and workspace, deletes the merged branch and closes the epic.

    line 45
  4. Distinctive:

    Only ACP agents. The script never falls back to a TUI agent.

    line 61

“Common” is measured: kept by at least 6 in 10 of the 172 edited versions of this file we could read. “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

Handoff

A handoff is a dated doc plus, when the repo uses beads, an epic. The doc is the brief. A session or agent starts only when the user says to start it. scripts/handoff does every mechanical step; the agent's job is to write a good doc and report what the script prints.

Commands

scripts/handoff new "Report importer"            # docs/YYYY-MM-DD-report-importer-handoff.md
scripts/handoff start DOC                        # epic + worktree + launch, detached
scripts/handoff start DOC --supervise            # wait for the agent, then collect
scripts/handoff start DOC --here                 # worktree, branch, brief and epic; no agent
scripts/handoff start DOC --agent "codex-acp"    # any ACP agent command
scripts/handoff status ID                        # commits, dirty files, agent exit, log tail
scripts/handoff collect ID                       # verify and record the result on the epic
scripts/handoff close ID [--discard]             # remove the worktree once merged
scripts/handoff list

Run them from inside the target repository. State is kept in <git-common-dir>/handoff/<id>/ (state.json, brief.md, run.log), so nothing is added to the working tree.

Workflow

  1. Write the doc. Use handoff new "<title>" or the project's own convention: look at existing handoff or spec docs first and match their folder and naming. Fill in problem, scope, what is out of scope and "done when". Put verification commands in a fenced block under a heading named Checks (or Verify): collect runs each line in the worktree.
  2. Stop and wait until the user says to start it. Writing the doc is not consent to launch.
  3. Start. handoff start DOC prints the base branch and commit it cuts from; relay that line, and any note that the source checkout has uncommitted changes the worktree won't have. Without bd or a .beads/ folder the epic is skipped with a one-line note.
  4. Check in with handoff status ID. Don't poll in a loop; use --supervise when the user wants to wait for it.
  5. Collect. handoff collect ID checks that the agent finished, the branch has commits, the worktree is clean, nothing new changed in the source checkout, and each brief check passes. It records the result as a comment on the epic and exits non-zero on any failure.
  6. Merge is the user's call. Report the verdict and the branch name, then stop.
  7. Close after the merge: handoff close ID removes the worktree and workspace, deletes the merged branch and closes the epic. It refuses an unmerged branch unless --discard, which removes the worktree but always keeps the branch.

What the agent is told

The doc is sent as the prompt with fixed rules appended: stay in the worktree, commit on the handoff branch with explicit paths only, never merge, rebase, push or open a PR, don't touch files outside the worktree, report blockers instead of working around them, and end with a short report. acp-run enforces the path rule for permission requests and file reads and writes.

Requirements and limits

  • git, Python 3, a running herdr, and herdr's acp-run script. The script looks for acp-run on PATH, next to this skill in a sibling herdr skill, and in the usual skill folders; set HANDOFF_ACP_RUN to override.
  • Only ACP agents. The script never falls back to a TUI agent.
  • The worktree is cut from committed history. Uncommitted changes, including an uncommitted doc, aren't in it; the brief still reaches the agent because it is passed as a file.
  • "Nothing changed outside the worktree" compares the source checkout's git status with a snapshot from start time. In a checkout shared with other sessions a change there may be theirs, so collect names the files instead of guessing.

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.