agentleFS
ArcanEdge-AIArcanEdge-AI/coding-agent-playbook-codex/skills/handoff/SKILL.md

handoff

Skill3 starsChanged 3 months agoMITView on GitHub

Start a new Codex chat with the complete material context from the current work when the user asks for a handoff, a fresh continuation chat, or everything worked on, researched, explored, validated, and verified.

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

10 rules. 0 are common, 10 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 172 edited versions we read keep these.

  • 10

    Distinctive

    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:

    Before composing the handoff:

    line 14
  2. Distinctive:

    Treat the completed conversation, repository, connected services, memory, and user statements as distinct evidence sources.

    line 16
  3. Distinctive:

    Read references/context-contract.md before assembling the context package.

    line 21
  4. Distinctive:

    Create a fresh chat with a self-contained context package.

    line 41
  5. Distinctive:

    If no thread-creation capability exists, provide one self-contained copy-paste prompt and say clearly that the chat was not created.

    line 48
  6. Distinctive:

    Do not claim success from a create response alone.

    line 52
  7. Distinctive:

    Read the new thread and confirm the context prompt is present.

    line 54
  8. Distinctive:

    Ensure the new prompt tells the next agent to recheck live state before acting.

    line 57
  9. Distinctive:

    Ensure no underlying merge, deployment, report launch, cloud mutation, publication, or destructive action started automatically.

    line 58
  10. Distinctive:

    Do not archive or delete the source thread unless the user separately asks.

    line 59

“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

Move the current work into a new chat without losing decisions, evidence, failures, workspace state, or the real next gate. When Codex app thread tools are available, actually create, verify, and open the new chat; do not stop at drafting a prompt.

The request to start a new chat authorizes thread creation, context seeding, titling, and navigation. It does not authorize code changes, merge, deployment, external publication, cloud mutation, destructive cleanup, or continuing the underlying project work.

Establish Current Truth

Before composing the handoff:

  1. Treat the completed conversation, repository, connected services, memory, and user statements as distinct evidence sources.
  2. Recheck cheap drift-prone facts that materially affect the next chat: repository path, branch/ref, commit, upstream, dirty state, PR/check/review status, deployment state, active worktree, and relevant external access state.
  3. Preserve the distinction between verified current, user-reported, historical, and unverified facts. Never upgrade source changes, a merge, or prior test output into deployed acceptance.
  4. Preserve unrelated local work. Record dirty or user-owned paths without cleaning, resetting, moving, staging, or overwriting them.

Read references/context-contract.md before assembling the context package.

Build the Context Package

Include all material context needed to continue correctly, not a transcript dump. Capture:

  • the objective and current verdict;
  • work completed and its exact artifacts;
  • research, experiments, alternatives, and causal findings;
  • validations run, exact outcomes, warnings, and evidence boundaries;
  • user decisions, approvals, rejected approaches, and non-negotiable invariants;
  • repository, worktree, branch, PR, release, and external-system anchors;
  • failures, dead ends, pre-existing defects, and why they were classified that way;
  • unresolved risks, missing evidence, and the precise next gate;
  • authorization boundaries and actions the new chat must not infer.

Use exact file paths, identifiers, commits, PRs, run IDs, test counts, and evidence locators when they materially prevent rediscovery or mistakes. Remove repetition and conversational noise. Never include credentials, tokens, private keys, session cookies, sign-in links, device codes, secret values, customer prose, or unnecessary sensitive payloads.

Create the New Chat

Create a fresh chat with a self-contained context package. A fork carries prior history and is not a fresh handoff. Use this order:

  1. Select the project safely: List projects first and match the current repository by exact path. Use the host's supported default environment. Create a worktree only when the user explicitly requests one and the project is a Git repository. Specify a starting state only when the user requests it: current working-tree state or a verified branch/ref. Otherwise use the host's default starting state. Never invent a branch.
  2. Create the thread: Seed a fresh project thread with the complete context package and a clear next gate.
  3. Handle asynchronous creation: A create action may return a clientThreadId while setup is pending. Never pass it to a tool that requires a real threadId. Poll/list threads and identify the new one from its requested title, project, path, creation time, and summary. Avoid creating a duplicate while setup is in progress.
  4. Title and open it: Give the thread a short, unambiguous title, then navigate the app to that thread.

If no thread-creation capability exists, provide one self-contained copy-paste prompt and say clearly that the chat was not created. Create a workspace handoff file only when the user requests one or when a durable artifact is necessary and the destination is safe.

Verify the Handoff

Do not claim success from a create response alone.

  • Read the new thread and confirm the context prompt is present.
  • Confirm the thread has a real threadId, expected workspace/project, and requested title.
  • Confirm it is active, ready, or has produced its acknowledgement.
  • Ensure the new prompt tells the next agent to recheck live state before acting.
  • Ensure no underlying merge, deployment, report launch, cloud mutation, publication, or destructive action started automatically.
  • Do not archive or delete the source thread unless the user separately asks.

Finish with a concise confirmation naming the new thread and its ID.

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.