agentleFS
jrgilbertsonjrgilbertson/the-rookery/skills/reviewing-meetings/SKILL.md

reviewing-meetings

Skill15 starsChanged 14 days agoMITView on GitHub

Use when the user asks to review, process, import, catch up on, or check for newly completed meetings from a configured meeting source, including Granola; when a scheduled post-meeting review invokes it; or when the user later revisits, resumes, approves, edits, defers, skips, or otherwise decides visible meeting-review actions, including CRM-derived actions. Do not use for preparing an upcoming meeting, taking live notes, maintaining a CRM, or directly creating tasks, issues, or calendar events without meeting review.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before choosing a source scope, determine whether the current user message only approves, edits, defers, skips, declines, revisits, or …

    line 18
  2. Distinctive:

    If the same message also explicitly asks for a new meeting check, finish the action-response phase first.

    line 31
  3. Distinctive:

    Use the disposition definitions and precedence in the source-interpretation reference.

    line 61
  4. Distinctive:

    Before preparing any newly proposed meeting, read references/action-routing.md and assets/review-bundle.md.

    line 71
  5. Distinctive:

    Keep the in-progress proposal ephemeral.

    line 79
  6. Distinctive:

    If current conversation history is unavailable, disclose that pending and reviewed or dismissed suppression cannot be verified.

    line 85
  7. Very specific:

    After refining the meeting synthesis, invoke managing-personal-crm in embedded mode when the evidence contains a possible relationship …

    line 95
  8. Distinctive:

    Follow the action-routing reference and review-bundle asset.

    line 112
  9. Distinctive:

    Only after synthesis, any embedded relationship evaluation, and action routing are complete, present the first and final visible review …

    line 122
  10. Distinctive:

    Nothing new: every discovered meeting was already pending, already approved, reviewed, or dismissed in this conversation.

    line 150
  11. Distinctive:

    Waiting for source: no meeting is ready, but at least one may become eligible when source processing or the meeting finishes.

    line 152
  12. Distinctive:

    When outcomes are mixed, use Partial if at least one useful proposal exists and another meeting or continuity check is limited.

    line 159

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “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

Reviewing Meetings

Turn completed meetings from the configured source into grounded proposals for review. Granola is one supported source, not part of the workflow's identity. Scheduled and manual entry use this same workflow. A schedule supplies timing, not extra authority: every run remains read-only until the user reviews a proposal.

Resolve visible action responses first

Before choosing a source scope, determine whether the current user message only approves, edits, defers, skips, declines, revisits, or otherwise decides actions from a visible meeting-review bundle. If so, read references/applying-approved-actions.md and handle that response directly. Do not rediscover meetings, run a broad meeting-source query, or append unrelated proposals. When application-time revalidation of an exact CRM-derived action requires evidence from its original meeting, the workflow may reread only the smallest necessary slice from the exact displayed source and native ID. This bounded reread is not discovery and must not produce another meeting or proposal. If the source read is unavailable or fails, or the exact source or native ID is ambiguous, make no write and report that action Manual.

If the same message also explicitly asks for a new meeting check, finish the action-response phase first. Then run discovery as a separate read-only phase; do not use newly discovered proposals to reinterpret the earlier action decision.

Completion: a response to visible actions was handled against that exact bundle before any separately requested discovery began.

Select the source scope

For a manual request, use the named meeting or time range. If neither is given, use the inclusive 168-hour interval ending at the query time. For a scheduled request, use the supplied range or the same 168-hour default. Compare boundary times as absolute instants. The overlap is intentional because eligibility comes from observable source state, not a last-run cursor.

Read references/source-interpretation.md before retrieving meeting data.

Completion: the run has a bounded source window and has loaded the source interpretation rules.

Discover and classify meetings

Retrieve meetings in the source scope, then classify every returned meeting from observable source state. Once a meeting is source-ready, use its source name and stable native ID as its exact identity. A waiting meeting may not have a stable ID yet. A meeting is eligible only when it has ended, has a stable ID, and contains enough source material for a grounded proposal.

Use the disposition definitions and precedence in the source-interpretation reference. Compare exact source names and IDs against approved notes and, when retrievable, visible proposal lifecycles and dismissals in this conversation.

Completion: every discovered meeting has one observable disposition, and no cursor, ledger, database, state file, transcript archive, or generated marker was created.

Synthesize only new proposals

Before preparing any newly proposed meeting, read references/action-routing.md and assets/review-bundle.md. Use the asset for every meeting proposal and the reference to resolve one authoritative destination for each supported action. Populate every applicable field in the asset, shape the preview with the configured live template, and leave unsupported details explicitly unresolved.

Keep the in-progress proposal ephemeral. Do not present it yet or create or edit a meeting note while preparing it. Later runs append only newly eligible meetings; they do not repeat, renumber, or recompute older pending or deferred proposals. A fully decided visible bundle remains suppressed by its conversational disposition rather than becoming a new proposal again.

If current conversation history is unavailable, disclose that pending and reviewed or dismissed suppression cannot be verified. Continue using exact source-and-ID checks against approved notes, and do not claim the result is free of conversational duplicates.

Completion: the in-progress synthesis contains every and only newly proposed meeting from this run, or establishes that there is no new meeting to review.

Evaluate relationship effects selectively

After refining the meeting synthesis, invoke managing-personal-crm in embedded mode when the evidence contains a possible relationship interaction, follow-up, durable effect, or useful person connection. That companion owns the relationship judgment. This workflow keeps the single meeting bundle, continuous action numbering, approval flow, and run ending.

Carry supported outputs into the existing meeting categories. A contextual connection may appear as a concise insight; every durable destination effect remains a separate numbered action. If the companion is unavailable, complete the meeting review and mark a specific otherwise-actionable relationship effect Manual only when its safe path is known to be unavailable.

Completion: relationship judgment adds only supported effects to the existing meeting proposal and never creates a second workflow or completion state.

Route and bind proposed actions

Follow the action-routing reference and review-bundle asset. They define the supported categories, canonical ownership checks, proposal fields, numbering, dependencies, and approval presentation. Omit any category whose evidence or destination is unresolved.

Completion: every proposed action is complete enough to approve independently, has one authoritative destination, and exposes every prerequisite.

Present one final review bundle

Only after synthesis, any embedded relationship evaluation, and action routing are complete, present the first and final visible review bundle. Include all newly proposed meetings from this run together, grouped by meeting, with one continuous action-number sequence. Carry every supported relationship effect into that bundle; do not expose a preliminary meeting-only bundle, nested CRM bundle, or any second proposal surface. A later run may add one terse count or reminder that older work remains pending.

Completion: one visible bundle contains the complete routed proposal, or the run states that there is no new meeting to review.

Apply only exact approvals

On a later user response, follow references/applying-approved-actions.md before writing. That reference owns approval binding, pre-write rereads, atomicity, dependency order, readback, retry boundaries, and application outcomes.

Completion: each decided action has one readback-backed outcome or an explicit safe stop, and no action was redirected, duplicated, or retried blindly.

End explicitly

End the read-only run as one of:

  • Ready for review: one or more new meeting proposals are waiting for the user.
  • Nothing new: every discovered meeting was already pending, already approved, reviewed, or dismissed in this conversation.
  • Waiting for source: no meeting is ready, but at least one may become eligible when source processing or the meeting finishes.
  • Partial: useful proposals were prepared, but a named source or continuity gap limits part of the result.
  • Unable to prepare: the run could not establish trustworthy meeting identity or content.

When outcomes are mixed, use Partial if at least one useful proposal exists and another meeting or continuity check is limited. With no useful proposal, prefer Waiting for source when every unresolved meeting may become ready later; use Unable to prepare when any collision or required check needs intervention. Use Nothing new only after a successful query with no new or unresolved meeting.

Summarize the count for each observed disposition. A scheduled run stops after presenting the read-only result and waits for the user; firing the schedule does not approve any durable change.

Completion: the ending matches the classified meetings, identifies any limit, and leaves every durable change unapplied.

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.