Capture

Skill nyechiel/ai-augmented-workflow/skills/capture

A template for building your own AI-powered knowledge worker workflow using Claude Code and MCP.

Install

npx -y skills add nyechiel/ai-augmented-workflow --skill capture

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Auto-discover unprocessed meeting notes and transcripts, capture ad-hoc thoughts and facts, or manually capture key facts from a session. Keeps domain knowledge files up to date.

SKILL.md

3.8 KB, as published. Nobody here has run it

Knowledge Capture

Auto-discover unprocessed meeting notes and shared documents, capture ad-hoc thoughts, or manually capture key facts into the right places in the repo.

Use user_google_email: "{YOUR_EMAIL}" for all Google Workspace calls.

Workflow

1. Determine Mode

Parse the argument:

No argument - scan calendar and inbox for meetings with uncaptured notes (last 24h)
Time window (e.g., 3d, 7d) - scan with that lookback period
Domain name (matches a directory with knowledge.md) - interactive capture for that domain
Meeting name(s) - search for that specific meeting's notes
Free-form text - classify and route immediately (URL, fact, action item, contact)

2. Auto-Discovery (no argument or time window)

Use get_current_time for today's date
Use get_events to list meetings in the time window
For each meeting, search Gmail for automated notes ("meeting notes" subject:"<title>")
If notes found, read the linked Google Doc via get_doc_as_markdown (not just the email summary)
Check meetings/ for existing capture files to avoid re-processing
Present discovered meetings and ask which to process

3. Process Meeting Content

For each meeting being processed, extract:

Action items assigned to you - create Crux tasks with label action-item
Decisions made - route to decision records if explicit choices were made
Key facts - route to domain knowledge
New contacts - look up via search_directory, add to stakeholders

4. Classify Each Fact

Every piece of knowledge goes to exactly one place:

Type	Where
Decision	`{domain}/decisions/<slug>.md` (use template)
Domain knowledge	`{domain}/knowledge.md`
Stakeholder context	`stakeholders/README.md`
Cross-session context	`memory/` files

5. Update Domain Knowledge

Append to {domain}/knowledge.md:

### [Topic] - [date]

[Fact, stated concisely. One paragraph max.]

**Source:** [Meeting brief link, doc link, or "verbal from [person]"]

Rules: append don't rewrite, one fact per entry, include source, deduplicate.

6. Generate Meeting Record

Save to meetings/YYYY-MM-DD-<topic-slug>.md using templates/meeting-record.md. If a prep brief already exists from /meeting-prep, update it instead of creating a new file.

7. Confirm Capture

Present a summary: what was captured, where it was saved, any Crux tasks created, any new stakeholders added.

Graceful Degradation

Scenario	Behavior
Calendar unavailable	Gmail-only discovery
Gmail unavailable	Calendar-only, scan Drive
Both unavailable	Fall back to interactive flow

Key Principles

Auto-discover first, ask second. Scan calendar and inbox rather than asking questions.
Facts over summaries. "Auth team owns SSO, target June 15" beats "We discussed SSO plans."
Source everything. If you can't trace it back, it's not knowledge.