agentleFS
ychampionychampion/ystack/skills/recall/SKILL.md

recall

Skill4 starsChanged 2 days agoMITView on GitHub

Rebuild where work on something stands from the repo, handoff notes and past agent sessions, check it against the live code, and return a short current-state brief. Use for "where did I leave off", "recall my work on X", "catch me up", "what was I doing in this repo", "status of X", or at the start of a session that continues earlier work.

145 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use addskill, apply-firmware 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.
  • ryanportfolioSkill

    addskill

    Create, import, update, or install repo or personal skills: runtime ownership, resources, discovery checks.

    Also in 6 projects that use this

  • ryanportfolioSkill

    apply-firmware

    Bring current folder (plain folder, existing repo, old or partial firmware copy) up to full Harness Firmware in place. Mirror external repo → adopt-repo.

    Also in 6 projects that use this

  • +10 more from ryanportfolio

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 132 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:

    After a break, a restart or a new session, the state of a piece of work is scattered: half in git, half in old conversations, some in a …

    line 12
  2. Distinctive:

    Keep it under 40 lines. If the user named a topic, leave out unrelated work in the same repo.

    line 67
  3. Distinctive:

    Don't resume the work on your own after the brief.

    line 73

“Common” is measured: kept by at least 6 in 10 of the 132 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

Recall

After a break, a restart or a new session, the state of a piece of work is scattered: half in git, half in old conversations, some in a handoff note. This skill gathers it, checks it against what's actually in the repo now, and hands back one brief.

Sources, cheapest first

  1. The repo, live. git status, git stash list, current branch, git log --oneline -15, branches touched in the last two weeks (git for-each-ref --sort=-committerdate refs/heads), and open pull requests (gh pr list --author @me when GitHub's CLI is set up).
  2. Notes the work left behind. Handoff files (ystack's context-meter writes them to ~/.claude/plans/<project>-handoff.md), TODO, CHANGELOG "Unreleased", docs/HANDOFF*, plan files in the repo.
  3. Past agent sessions. Claude Code keeps transcripts in ~/.claude/projects/<slug>/*.jsonl, where <slug> is the working directory with separators replaced by dashes. Codex keeps ~/.codex/sessions/. Other agents keep their own logs; find them by modification time.

Reading transcripts without flooding the context

Transcripts are large. Never read one whole. Pick the two or three most recent that mention the topic, then pull out only:

  • the user's prompts (what was asked, in their words);
  • the last few long assistant messages (what was reported at the end);
  • the last lines mentioning a test, build or push result.

A short script that streams the JSONL and prints those parts, each truncated to a few hundred characters, is the right tool. If you're delegating, have a cheap subagent extract and quote them; you do the judging.

Check before you trust

A transcript records what an agent said, not what's true now. Before the brief states something:

  • "done" or "merged": confirm the commit or PR exists and is where it was said to be;
  • "tests pass": run them, or mark the claim as from the transcript and unverified;
  • a file was changed: confirm the change is in the working tree or a commit;
  • a background job or agent was running: assume it's gone unless you can see it.

Mark anything you couldn't confirm. Old state presented as current is worse than no state.

The brief

# <topic>: where it stands (<date>)

Goal: <one line, in the user's words if possible>
Done (verified): <items, each with the commit, PR or command that shows it>
In flight: <uncommitted changes, open PRs, branches, with paths>
Broken or failing: <errors and failing tests, quoted>
Unverified (from past sessions only): <claims you couldn't confirm>
Waiting on the user: <decisions, approvals, logins>
Next step: <the one concrete thing to do first>
Sources: <the transcripts, notes and commands this came from>

Keep it under 40 lines. If the user named a topic, leave out unrelated work in the same repo.

Rules

  • Live repo beats notes, and notes beat transcripts, when they disagree. Say which one you trusted.
  • Quote errors word for word; don't paraphrase them.
  • Don't resume the work on your own after the brief. Show it and let the user pick the next step, unless they asked you to carry on.

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.