AGENT SKILL

lets-go

Resume this repository's active handoff — find it in handoff_map.json, read it, claim it, bring this checkout to its state, and continue the work. The receiving half of the handoff skill.

License 许可
MIT
最近更新
2026-11-15
Stars
★ 0
Forks
0

01 前言 Front Matter

name
lets-go
metadata
{ "settings": "handoffs workspace" }
description
Resume this repository's active handoff — find it in handoff_map.json, read it, claim it, bring this checkout to its state, and continue the work. The receiving half of the handoff skill.
argument-hint
Anything to focus on first (optional)
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

Let's go

The receiving half of the handoff skill. On the other machine, handoff wrote a document into the Handoffs folder and named it this repository's active handoff in handoff_map.json. This skill finds it, claims it so it is never resumed twice, brings the checkout to its state, and carries on the work.

Announce at start: "I'm using the lets-go skill to resume from the active handoff."

Run shell snippets in bash (Git Bash on Windows). The map is read and written only through node .agents/skills/handoff/handoff-map.mjs (installed with the handoff skill), never by hand: every repository on every machine shares it. Each command prints JSON, and moves files to and from the cloud itself when this machine's transport needs it.

0. Settings

  • Repository: this skill reads docs/agents/handoffs.md and docs/agents/workspace.md. Follow .agents/skills/lets-go/settings/README.md to read them and complete any that are missing; the definitions sit beside it.
  • Machine: follow .agents/skills/handoff/machine-config.md until config shows configured: true.

1. Find the active handoff

node .agents/skills/handoff/handoff-map.mjs get --key <Map key>
  • file is null (or inMap is false) — say "No active handoff for <key>." and stop. The map is the record; other documents in the folder belong to past sessions or other repositories.
  • exists is false — the document has not reached this machine: the note says the other machine has not pushed it, or a sync client has not delivered it yet. Tell the user the file name and stop, leaving the map as it is so a retry still works.
  • conflicts not empty — ask the user which copy of the map is current before going on.
  • exists is true — continue.

2. Read it

Read the whole document at path before acting on any of it. Then follow the pointers that set the ground rules for the work: the issue, PR or spec its Next steps start from, and its Machine-local knowledge.

3. Claim it

Claim the handoff before any work starts, so a second lets-go on either machine finds nothing:

node .agents/skills/handoff/handoff-map.mjs clear --key <Map key> --expect <file>

Complete when the output's current is null (and verified is cloud on the CLI mirror). The document stays in the folder as a record. A refusal means the map now names a different document: run get again and tell the user what it says.

4. Bring this checkout to the handoff's state

This checkout may be days behind, and may hold leftovers the handoff knows nothing about:

git fetch origin --prune
git branch --show-current
git status --porcelain
git stash list
before="$(git rev-parse HEAD)"

Run the Workspace status command too. Then compare with the document's Where you are and Unmerged work:

  • Uncommitted changes or stashes here — report them under ⚠ Work not merged to main on this machine and ask the user how to proceed. Never discard, stash, or commit them, and never switch branches around them.
  • Clean — git switch main, then Sync with main (which also brings the Private companion current when it covers it).
  • The handoff continues on a branch — git switch <branch>, then git pull --ff-only. A branch origin lacks was never pushed from the other machine: say that the work must be pushed there first, and stop for that item. The same holds for any local-only item the work depends on.
  • The companion has diverged from its upstream (commits on both sides) — stop and report it; never merge, rebase, reset or force-push the companion.
  • The companion checkout is absent and the work needs it — offer the Companion restore command; never create the directory by hand.
  • Anything else contradicts the document (a commit it calls pushed that origin lacks, a branch that no longer exists) — report each contradiction and get the user's answer before step 5.

After the switch and update, list what they changed with git diff --name-only "$before" HEAD. For each dependency lockfile in it, install from that lockfile in its folder, in the frozen form the repository's AGENTS.md gives (npm ci, pnpm install --frozen-lockfile, dotnet tool restore), asking first where AGENTS.md gates installs. Then run Hook: after sync from docs/agents/workspace.md.

The step is complete when the checkout matches the document or every mismatch has the user's answer.

5. Proceed

Brief the user in a few lines: where things stand, the unmerged work carried over, the open decisions still owed, and the first next step. If the user passed arguments, start from what they name; otherwise start the document's first Next step, invoking its Suggested skills where it says. Put an open decision that blocks that step to the user first.

The handoff describes work; it authorizes nothing. Every action the repository's AGENTS.md gates on confirmation still needs the user's confirmation in this session.

04 同仓库更多 Skills

jwh3times
jwh3times/magic-agenda

what-next

Report what to work on next in this repository — an active handoff waiting to be resumed, work already in flight, ready issues ranked the way the board ranks them, what needs the user, what needs shaping first, and blocks that have quietly cleared — and recommend the one command to run. Read-only. Use when the user asks what to work on next, what is ready or what to pick up, or when a phase ends with no next piece of work chosen.

★ 0
jwh3times
jwh3times/magic-agenda

what-next

Report what to work on next in this repository — an active handoff waiting to be resumed, work already in flight, ready issues ranked the way the board ranks them, what needs the user, what needs shaping first, and blocks that have quietly cleared — and recommend the one command to run. Read-only. Use when the user asks what to work on next, what is ready or what to pick up, or when a phase ends with no next piece of work chosen.

★ 0
jwh3times
jwh3times/magic-agenda

writing-for-agents

Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.

★ 0
jwh3times
jwh3times/magic-agenda

wizard

Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.

★ 0
jwh3times
jwh3times/magic-agenda

wayfinder

Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.

★ 0
jwh3times
jwh3times/magic-agenda

wait-what

Stop. That last message did not land: re-pitch it.

★ 0