agentleFS
mikel-codemikel-code/claude-skills/granola/SKILL.md

granola

Skill0 starsChanged 3 months agoMITView on GitHub

Access Granola meeting notes through the Granola API with a leak-resistant Proton Pass key (pass-cli, dedicated-vault agent). List meeting notes, show a note's summary, attendees, and optionally its transcript, and export notes to Markdown files. Use for /granola, \"list my granola meetings/notes\", \"show the notes/summary/transcript of that meeting\", or \"export that meeting to markdown\".

  • Reads credentials

10 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

5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read meeting notes from the Granola API without the API key ever entering your context.

    line 10
  2. Very specific:

    Never: read, print, or export the API key: no pass-cli item view, no pass-cli run -- env, no echo $GRANOLA_API_KEY, no copying it into …

    line 26
  3. Very specific:

    Never wrap your own process (or any long-lived process) in pass-cli run — one launcher invocation per command, carrying only this key.

    line 30
  4. Very specific:

    Run everything as <skill-dir>/scripts/granola.sh <subcommand>:

    line 45
  5. Very specific:

    Create the granola-reader agent, vault-scoped to that dedicated vault — see the runbook for exact commands.

    line 102

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

Granola

Read meeting notes from the Granola API without the API key ever entering your context. All API access goes through scripts/granola.sh, which reads a pass:// reference from granola.env and resolves it via pass-cli run, which injects the real key into the client process only.

The reference lives in granola.env, the native dotenv format pass-cli run --env-file consumes directly — not secrets.toml. This skill intentionally diverges from the TOML-registry pattern used elsewhere (e.g. the second-brain .tooling/secrets.toml): there is no cross-vendor standard for ENV_VAR → secret-reference mappings, and dotenv-with-references is the closest thing to a de-facto one (it's what pass-cli and 1Password's op both implement, flag-for-flag identical --env-file semantics).

Security rules — non-negotiable

  • Never read, print, or export the API key: no pass-cli item view, no pass-cli run -- env, no echo $GRANOLA_API_KEY, no copying it into files.
  • Only access the API through scripts/granola.sh. Do not hand-craft curl calls to public-api.granola.ai.
  • Never wrap your own process (or any long-lived process) in pass-cli run — one launcher invocation per command, carrying only this key.
  • pass-cli run's output masking is best-effort: it only catches literal occurrences of the resolved value in streams it controls. An encoded/ transformed key, or anything the client writes straight to a file, is not covered — don't rely on masking as a guarantee.
  • Direct pass-cli invocation from a Claude Code session may be hook-blocked at the harness level (session repointed to a dead end, plus a PreToolUse deny on any pass-cli/PROTON_PASS_ command) — this reinforces, not replaces, the "only through scripts/granola.sh" rule above. If it fires, it means something tried to reach pass-cli directly; route through the script instead of working around the block.

Commands

Run everything as <skill-dir>/scripts/granola.sh <subcommand>:

granola.sh list [--limit N] [--all] [--after ISO] [--before ISO]
                [--updated-after ISO] [--folder fol_ID] [--cursor CURSOR]
granola.sh folders
granola.sh show <not_ID> [--transcript]
granola.sh export <not_ID> [--transcript] [-o PATH]
granola.sh sync --dir PATH [--transcripts] [--folder fol_ID] [--full]
  • list prints one note per line: not_ID created-at title — owner. Default page size 20; if more exist it prints the --cursor to continue with, or use --all to follow pagination to the end.
  • show prints the note as Markdown to stdout: title, date, organiser, attendees, folders, Granola web link, and the AI summary. Add --transcript to append the full transcript (speaker-attributed, timestamped). Only fetch the transcript when the user asks for it — transcripts are large.
  • export writes that same Markdown to a file (default ./<date>-<title-slug>.md in the current directory) and prints the absolute path. Use -o when the user names a destination. Exported files start with an Open Knowledge Format (OKF) v0.2 YAML frontmatter block (type/title/description/resource/tags/generated/sources plus a meeting: extension with attendees and organiser).
  • sync incrementally mirrors notes into --dir as OKF-frontmatter Markdown files: first run (or --full) fetches everything; later runs fetch only notes changed since the last sync (tracked in <dir>/.granola-sync.json, which also pins each note id to its filename — don't edit or delete it). Sync never deletes local files: notes removed upstream are reported as "kept local file" (only detectable on --full runs) and left in place; it also never overwrites files it doesn't manage. Use this from projects that keep a meeting-notes mirror — point the project's CLAUDE.md at granola.sh sync --dir <notes-dir> and treat the mirrored files as read-only source material.

Typical flows: "what meetings did I have this week" → list --after <monday>; "show me the notes from the X meeting" → list to find the id, then show; "export it with the transcript" → export <id> --transcript.

Rate limit is 5 requests/second — fine for normal use; avoid tight loops over show.

Setup (one-time, human present)

The service ships unconfigured. granola.sh prints a short version of these steps itself when granola.env is missing; full commands, including creating and renewing the Proton Pass agent token, are in references/proton-pass-setup.md — the user must run all of it themselves, never on the agent's behalf, since it handles a raw credential.

  1. In Granola (web app → Settings → API keys), create an API key (grn_…).
  2. In Proton Pass, create a dedicated Granola vault and store the key there as item Granola, custom field named exactly API Key (field names are case-sensitive; a dedicated vault is required, not just an item-level grant — see the runbook for why).
  3. Create the granola-reader agent, vault-scoped to that dedicated vault — see the runbook for exact commands.
  4. Copy granola.env.example to granola.env in this skill directory (the reference is already correct; nothing to edit).

Troubleshooting

  • "not configured" / granola.env missing → setup above not completed; walk the user through it.
  • "runtime session expired and the stored agent token could not be fetched" (exit 3) → granola.sh self-heals the ~2h runtime session automatically on every invocation; this only surfaces when that self-heal itself failed — either the user's personal Proton Pass session is stale (ask them to run pass-cli login normally) or the token was never stored (point them at the runbook's "Store the token" step). Never attempt the login for them — it requires a token only they hold.
  • "found the stored agent token but could not log the runtime session in" (exit 3) → the agent token itself expired (yearly, not the ~2h session). Point the user at the runbook's "Token expired" section; never attempt the renewal for them.
  • 401 from Granola → key missing/revoked in the Proton Pass item; ask the user to check item Granola, field API Key, in the dedicated Granola vault.
  • pass-cli not found → install Proton Pass CLI, or set PASS_CLI to its absolute path (cron/scheduled runs have minimal PATH).

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.