agentleFS
bendyworksbendyworks/claude-skills/skills/resume-work/SKILL.md

resume-work

Skill1 starsChanged 39 days agoMITView on GitHub

Carry on an unfinished story in a fresh session. Reads the story's plan file and its resume notes, checks them against the branch, the remote, and the pull requests before trusting them, rebuilds the task list under the plan's own numbers, says what is waiting on the developer, and starts the next to-do. Use when the user says "resume <name>", "pick <story> back up", "continue where the last session left off", or invokes the resume-work skill. Not Claude Code's own /resume, which reopens an old conversation.

5 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

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

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 3

    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:

    When no plan file matches, list the plans and ask.

    line 20
  2. Distinctive:

    Check before trusting: Run:

    line 28
  3. Distinctive:

    Stop and ask: when a pull request for the story has merged or closed and the notes do not say so, when the branch's own remote branch holds …

    line 43

“Common” is measured: kept by at least 6 in 10 of the 5 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

Resume work

What the last session learned survives in the plan file: its to-dos, and the resume notes the pause-work skill (bundled in this plugin) appends. Notes describe the moment they were written; a merge or push since then makes them wrong with no error.

Steps

  1. Find the plan. The name given is the plan's filename without .md; with no name, use the current branch's. The plan-issue skill (bundled in this plugin) keeps plans at .work/plans/<name>.md. Look there first, then at .claude/plans/<name>.md for a plan made before .work/ existed. A project may keep them elsewhere. When no plan file matches, list the plans and ask.
  2. Read the plan whole, and any file its last ## Resume notes section points at. Act on that section only: what earlier ones list as waiting may since be settled. The story's branch is the one those notes name; without notes, the one named like the plan.
  3. Check before trusting. Run:

    git fetch <remote>
    git branch --show-current
    git status --short
    git log --oneline <remote>/<branch>..   # local commits not on the remote
    git log --oneline ..<remote>/<branch>   # remote commits not here
    git stash list
    gh pr list --head <branch> --state all --json number,title,state,isDraft

    Say where the plan and its notes disagree with what these show, before anything else. Then:

    • Stop and ask when a pull request for the story has merged or closed and the notes do not say so, when the branch's own remote branch holds commits this checkout lacks, or when the story's branch is missing or is not the one checked out. Change nothing first: no switch, pull, or stash.
    • Any other difference (a commit the notes call local is now pushed, a stash is gone) is reported and stops nothing: go on to step 4 with what is true now. What the notes say is waiting on the developer is still waiting.

    <remote> is the project's remote, usually origin; never use @{u}, which may be unset or name the base branch. With no <remote>/<branch>, the branch is not on the remote: no reason to stop.

    With no resume notes in the plan, say so, and say what cannot then be known: the last session's decisions and their reasons, which tests fail on purpose, and what it was waiting on.

    Uncommitted changes, and stashes made on the story's branch, that the notes do not account for are unexplained work: describe them and ask before building on them or discarding them.

  4. Rebuild the task list from the plan's unchecked to-dos, in the session's task list and not as a list in chat: one task for each, in order, the plan's own number leading the subject (4. Add dedupe_slug).
  5. Say what is waiting on the developer, each item with its text or the path to it.
  6. Start the next to-do, under the plan's own rules for working one. A to-do the notes show waiting on the developer (a review, an answer, an approval) is never started on a guess: ask for what it waits on and stop there, with nothing written toward it.

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.