Skip to content
Agentmart
Submit a tool

remember

Remember a task across sessions in project-local state.

Skill by grugh

IndexedMIT
Write a review

Chalk Labs

Need help deploying remember?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/grugh-remember

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Remember a task across sessions in project-local state. Explicitly invoke to load a task or save its state; the first save sets up the project.

Use cases: productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/remember

What's inside

Skill name
remember
Description
Remember a task across sessions in project-local state. Explicitly invoke to load a task or save its state; the first save sets up the project.
License
MIT
Metadata
source: Lauren Tan, pstack: session pickup and recall; Matt Pocock, skills: context persistence. · opencode/autoinvoke: false
Instructions
76 lines in SKILL.md · view on GitHub

Files (4)

  • LICENSE
  • SKILL.md
  • openai.yaml · agents/
  • project-state.md · assets/

SKILL.md (first lines)

# Remember

There are no subcommands. Find the task, then pick the action from the session:

- **Load:** the task isn't loaded in this session yet and a match exists. Read it
  and report. If the request asks to save, run Save right after loading.
- **Save:** the task was already loaded or created in this session. Write
  `state.md`.
- **New:** nothing matches and the user describes work worth tracking. Set up if
  needed and create the task. If nothing matches and the request is vague, ask.

Free text after the command steers the result ("just summarize", "is it stale?").
If the user says not to use memory, or to only read it, obey for the session.

Read [the protocol](assets/project-state.md) on every invocation. It lives only in
this skill, so a project never holds a stale copy.

`state.md` is the only status record. Remember never reads or writes the
repository's docs; they change only when the user asks for a specific edit.
Progress, pending work and next steps never go into tracked files.

## Setup (every write)

1. Use the Git worktree root. Keep all writes inside `.local/agents/`; don't follow
   symlinks out of it.
2. If `git ls-files -- .local/agents/` lists anything, report it and stop. Never
   untrack files.
3. If `.local/agents/.gitignore` is missing, create it containing only `*`, so the
   folder ignores itself. Never edit the repository's ignore files. Then run
   `git check-ignore -q .local/agents/README.md`; if it fails, report it and stop.
4. If `.local/agents/README.md` is missing, create it with the text below. Never
   change an existing one; the skill doesn't read it.

   ```markdown
   # Agent task state

   Local task notes written by the remember skill. Never committed.
   The format lives in the skill: invoke remember to load or save a task.
   ```

## Find the task

Match loosely, the way the user would describe it. List `tasks/` and `archive/`,
and skim each candidate's **Goal**. Weigh the request's words, the current branch,
recent commits and this session's work. A bare invocation with one active task,
or one matching the branch, means that task.

Use a clear best match. If two or more are plausible, list them with their goals
and ask. Continue an archived task in place.

## Load

1. Read `state.md`. Compare its saved commit with `HEAD`, and check `git status`
   and the branch.
2. Report the goal, status, open issues and next step, and call out anything stale.
   If the task is in an older format, say that the next save converts it.
   Saved notes are context, not permission to run the next step.

## Save

1. Check the branch, `git status` and the relevant diff, and correct stale notes.
2. Write `state.md` in the protocol's current format, overwriting in place. Convert
   an older task the same way: keep its old files, link them from `state.md`, and
   move reference material out of `state.md` into topic files. A new task
   goes in `tasks/<name>/state.md`: the ticket as written, if there is one, plus a
   short kebab-case topic. Never invent a ticket.

After loading or saving, keep `state.md` current at milestones for the rest of
the session, as the protocol describes. If the user says "the docs" and could
mean either `state.md` or the repository's docs, ask once. Never commit or
archive automatically.

Return the path, a one-line status, the next action and any unverified gap.

From grugh/skills, MIT.

More from grugh/skills

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 22, 2026 (9 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from grugh

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get remember running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying remember?

CallGet help
Cite this page

remember skill. Agent Mart. https://agentmart.dev/skills/grugh-remember (last updated 2026-10-24)