agentleFS
jwm-axonijwm-axoni/claude-workbench/skills/capture/SKILL.md

capture

Skill2 starsChanged 5 days agoMITView on GitHub

Write back from a chat into the workbench the only way a chat is allowed to - one line appended to Tracking/Captures/YYYY-MM-DD.md per decision, process change, fact, correction or briefing answer settled in the conversation, plus a copy of every draft handed over to send, saved to Tracking/Drafts/ so the recap can compare it with what was actually sent. The nightly dream files these into the right pages. Works from any chat in any surface (chat, Cowork, Claude Code), not only inside the workbench project. Use whenever a chat settles something worth remembering, when the user corrects Claude or says "always"/"never", or says "remember this", "capture that", "log this decision", or /workbench:capture.

57 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use weekly, ingest and 8 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.
  • gaiaaiagentSkill

    weekly

    Generate a weekly digest rolling up dailies into higher-level pattern analysis

    Also in 4 projects that use this

  • JaimeGBASkill

    ingest

    Digest an external source (article, paper, transcript, URL, book chapter) and integrate its content into the knowledge base

    Also in 8 projects that use this

  • AnthropicSkill

    brief

    Generate contextual briefings for legal work — daily summary, topic research, or incident response. Use when starting your day and need a scan of legal-relevant items across email, calendar, and contracts, when researching a specific legal question across internal sources, or when a developing situation (data breach, litigation threat, regulatory inquiry) needs rapid context.

    Also in 6 projects that use this

  • dubwoahSkill

    ask

    Answer questions about the codebase without generating or modifying any code. Use when the user wants to understand how something works, find where something is defined, or explore the project structure. Invoked with /ask.

    Also in 4 projects that use this

  • termdockSkill

    recall

    Retrieves cross-session memories about past decisions, lessons, and patterns. Use when user asks about prior work, past decisions, or mentions a feature. Triggers include check memory, what did we decide, how did we solve.

    Also in 6 projects that use this

  • AlexPEClubSkill

    refine

    Always use when the user wants to discuss an exsiting feature or specification. Open an existing feature spec to improve, extend, or fundamentally challenge it. Pass the feature ID as argument (e.g. /refine PROJ-2).

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 55 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Don't capture chit-chat, drafts, half-formed ideas, or anything already in the vault and unchanged.

    line 37
  2. Distinctive:

    If the owner says "don't save this" or the topic is personal, don't.

    line 38
  3. Distinctive:

    Write every line in your own words as a summary of what was settled.

    line 50
  4. Very specific:

    When a chat establishes a relationship ("Sam's taking over the Acme account", "Dana reports to Lee now"), capture it as a fact: so the edge …

    line 58
  5. Very specific:

    If a line supersedes an earlier capture, say so: (supersedes 10:40 line about launch date).

    line 63
  6. Distinctive:

    If the file doesn't exist yet, create it with this frontmatter first:

    line 65
  7. Distinctive:

    If you revise it in the same chat, save the final version under the same name (it's the one they'll send).

    line 96
  8. Distinctive:

    Never capture passwords, keys, tokens, account numbers, or anything the role guardrails in the root CLAUDE.md exclude (e.g.

    line 102

“Common” is measured: kept by at least 6 in 10 of the 55 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

Capture

Chats are invisible to the routines. Captures are how what's settled in a chat reaches the vault: you append one line per item; the nightly dream files it into the right page and the weekly flags anything left unfiled.

This works from any chat, in any surface: the workbench project, another Cowork chat, a Claude Code session, a plain chat. The chats that settle things outside the project are exactly the ones that otherwise write nothing down. Wherever you are, the destination is the same: the workbench's Tracking/Captures/ folder, same file names, same line shape. Find the workbench the way the commands do: the nearest folder above the current one containing .workbench/manifest.json, or the folder the project is attached to. Set source: to where the chat happened (project-chat, cowork-chat, code-session, chat).

If this chat can't reach the workbench folder, you can't file the capture. Write the formatted capture lines in your reply instead, say plainly that they weren't saved and why, and ask the owner to paste them into the project (or today's capture file) so the dream picks them up.

You write two places: Tracking/Captures/<today>.md and Tracking/Drafts/ (today in the owner's timezone, from Context/me.md). Append only. Never edit or delete an earlier line, never write any other vault file from a chat. Everything else has an owner, and a second writer causes conflicts.

When to capture

Without being asked, at the moment it happens:

  • a decision gets made ("let's go with vendor B", "we're pushing launch to March")
  • a process changes ("from now on, QBR decks go through Dana first")
  • a fact comes up that the next session should know (a new owner, a date, a number, a name)
  • the owner corrects you, or says "always" or "never". Capture before continuing.
  • the owner answers a numbered question from a briefing ("3 yes, 5 it's Dana"). One answer: line per answer.

Don't capture chit-chat, drafts, half-formed ideas, or anything already in the vault and unchanged. If the owner says "don't save this" or the topic is personal, don't.

Line format

- decision: <what was decided, and why if stated> — about [[Page]] `source: project-chat` `date: YYYY-MM-DD`
- process: <what changed in how something is done> — about [[Page]] `source: project-chat` `date: YYYY-MM-DD`
- fact: <the fact, with who said it if relevant> — about [[Page]] `source: project-chat` `date: YYYY-MM-DD`
- correction: <what was wrong> → <the rule going forward> `source: project-chat` `date: YYYY-MM-DD`
- answer: briefing YYYY-MM-DD Q<n> — <the answer> — about [[Page]] `source: project-chat` `date: YYYY-MM-DD`

Write every line in your own words as a summary of what was settled. Never paste text from an email, document, message or web page into a capture. Something discussed but not decided is not a capture.

  • One item per line, in plain words a stranger could act on. Name people in full, as [[wikilinks]] when they have a page, and name every org, project and tool involved the same way. The dream builds relationship edges from these names, so "Pat said Acme wants it" is weaker than "[[Pat Lee]] said [[Acme Data]] wants the pilot moved — about [[Acme Expansion]]".
  • When a chat establishes a relationship ("Sam's taking over the Acme account", "Dana reports to Lee now"), capture it as a fact: so the edge gets built.
  • about [[Page]] names the page it belongs on: a project, person, org, tool, or Context/me / Context/voice for how the owner works or writes. Use the page's exact name; check it exists. If there's no page yet, use the plain name without brackets; the dream decides whether to create one.
  • If a line supersedes an earlier capture, say so: (supersedes 10:40 line about launch date).

If the file doesn't exist yet, create it with this frontmatter first:

---
type: reference
tags: [tracking, captures]
date: YYYY-MM-DD
---

Then tell the owner in a few words what you captured ("Captured: decision to push launch to March"). No ceremony.

Drafts

Separately from captures: whenever you hand the owner a draft they'll send to someone else (email, Slack message, post, doc section), save it exactly as given to Tracking/Drafts/YYYY-MM-DD_<short-slug>.md:

---
type: draft
date: YYYY-MM-DD
channel: email | slack | doc | post | other
to: "[[Name]]"
about: "[[Page]]"
---

<the draft exactly as given>

If you revise it in the same chat, save the final version under the same name (it's the one they'll send). Don't announce it. The recap compares it with what was actually sent, and the dream learns the owner's voice from the differences. Never save personal or HR drafts.

Secrets and sensitive data

Never capture passwords, keys, tokens, account numbers, or anything the role guardrails in the root CLAUDE.md exclude (e.g. candidate or employee personal data). Capture that the thing exists and where it lives, not its value.

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.