agentleFS
link7373link7373/agentic-bi-team/.claude/skills/triage/SKILL.md

triage

Skill4 starsChanged 7 days agoMITView on GitHub

Classify, size, route, and log an incoming BI request — decide what kind of work it is, what decision it serves, who on the team does it, and whether it should be done at all. Use as the front door for any request that isn't obviously one thing, and for anything arriving from outside the team.

190 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use writing-beats, setup-matt-pocock-skills and 10 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.
  • LesaniSkill

    writing-beats

    Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.

    Also in 74 projects that use this

  • GoogleSkill

    setup-matt-pocock-skills

    Configure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.

    Also in 182 projects that use this

  • SethGammonSkill

    ascii-diagram

    Generate perfectly aligned ASCII diagrams — architecture, flow, sequence, box-and-arrow. Uses a programmatic character-grid approach so alignment is guaranteed by math, not token prediction. Includes post-render verification.

    Also in 4 projects that use this

  • simo97Skill

    migrate-to-shoehorn

    Migrate test files from `as` type assertions to @total-typescript/shoehorn. Use when user mentions shoehorn, wants to replace `as` in tests, or needs partial test data.

    Also in 74 projects that use this

  • LesaniSkill

    writing-shape

    Take a markdown file of raw material and shape it into an article through a conversational session — drafting candidate openings, growing the piece paragraph by paragraph, arguing about format (lists, tables, callouts, quotes) at each step. Use when the user has a pile of notes, fragments, or a rough draft and wants help turning it into something publishable.

    Also in 70 projects that use this

  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 47 projects that use this

  • +1 more from Lesani

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 179 edited versions we read keep these.

  • 3

    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:

    Use it when a request is vague, large, arriving from outside the team, or one you suspect has been answered before.

    line 14
  2. Distinctive:

    Ask the one question that matters: what will you do differently depending on the answer?

    line 22
  3. Distinctive:

    If there's a clear answer, everything else gets easier — you know when to stop, what precision is enough, and what to lead with.

    line 25
  4. Very specific:

    Add a row to knowledge/request-log.md: date, requester, the request, the decision it serves, type, size, routed-to, and status.

    line 101
  5. Very specific:

    Every month or so, read knowledge/request-log.md as a dataset rather than a list:

    line 113

“Common” is measured: kept by at least 6 in 10 of the 179 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

Triage — The Front Door

Real BI teams don't fail because the analysis is bad. They fail because forty undifferentiated requests arrive, everything is urgent, nothing is written down, and the same question gets answered three times by three people with three different numbers.

This skill is the front door. It costs two minutes and routinely saves a day.

Use it when a request is vague, large, arriving from outside the team, or one you suspect has been answered before. Skip it for something you can answer in one query — running triage on "how many customers signed up yesterday?" is its own kind of waste.

Procedure

1. Find the decision

Ask the one question that matters: what will you do differently depending on the answer?

If there's a clear answer, everything else gets easier — you know when to stop, what precision is enough, and what to lead with. If there genuinely isn't one, that's the most valuable thing triage produces: some requests are curiosity (fine, but they queue behind decisions) and some are a reflex nobody has questioned since a reorg. Say so kindly and ask whether it's still needed.

Also establish when the decision gets made. A perfect answer after the meeting is worth nothing, and a rough answer with stated uncertainty before it is worth a lot. If the deadline forces a shortcut, agree the shortcut now rather than discovering it later.

2. Check whether it's already answered

Before classifying, look:

  • analyses/README.md — has this question been investigated?
  • knowledge/metrics-catalog.md — is the metric already defined? (If the request implies a new definition for an existing metric, stop and route to metrics-steward.)
  • dashboards/README.md — does a dashboard already show this?
  • knowledge/request-log.md — has someone asked this recently?

A request that has been asked three times is not a fourth analysis. It's a dashboard or a mart, and saying so is a better answer than doing the work again.

3. Classify

TypeLooks likeRoutes to
Question"how many", "why did", "which segment"bi-analyst → /analyze
Incidenta number looks impossible, a report is blank, a pipeline faileddata-quality-engineer → knowledge/incident-runbook.md
Metric disputetwo sources disagree, "what counts as active?"metrics-steward → /define-kpis
Recurring view"can I see this every week"dashboard-developer → /build-dashboard
New data"can we analyse X" where X isn't in the warehousedata-engineer → /build-pipeline
Performance"this query takes forever", "the dashboard is slow"analytics-engineer
Prediction"who will churn", "which accounts will upgrade"data-scientist → /build-model
Forecast"what will revenue be next quarter", "how much capacity will we need"data-scientist → /forecast
Impact / causal"did the price change cause the churn", "what was the launch worth"data-scientist → /causal-analysis (or /experiment if the change has not happened yet)
Privacy / permission"are we allowed to use this data", a new source with personal data, an access or deletion requestgovernance-officer → /privacy-review
Is the team right?"that number was wrong", "can we trust these answers"orchestrator → /eval (a correction becomes /eval add)
Deliverable"I need this for the board on Thursday"insights-communicator → /make-deliverable

Two classification traps worth naming, because both are common:

  • An incident dressed as a question. "Why did revenue drop 40% yesterday?" is usually a data incident, not an analysis. Check freshness before commissioning a deep dive — about half of these are a stale table, and the analysis time is wasted.
  • A dashboard request that's really one question. "Can I get a dashboard for X" often means "I need to know X once." Ask whether they'll look at it again next week. A dashboard nobody opens costs refresh, maintenance, and a chance to be wrong forever.

4. Size it

Rough and honest beats precise and late:

  • S — under an hour. One query against known tables.
  • M — under a day. Multiple sources, some new logic, a written finding.
  • L — multiple days. New pipeline, new model, a dashboard, or cross-source joins with unverified keys.
  • XL — over a week, or blocked on access, data that doesn't exist, or a decision nobody has made.

For anything L or larger, offer the S version first. A rough cut today usually answers the actual decision, and if it doesn't, it tells you exactly what the full version needs to cover. Say what the cheap version would and wouldn't tell them.

5. Prioritise honestly

Rank against what's already in flight, using: the size of the decision, whether the deadline is real, whether anything is blocked on it, and how much it costs.

Then say what it displaces. "Yes, by Thursday" is not a commitment unless you also say what moves. Everything-is-priority-one is how a BI team becomes a queue nobody trusts — and the standing cadence (scorecards, health checks) does not get displaced by ad-hoc work, because it's the thing that catches problems nobody asked about.

6. Log it and hand off

Add a row to knowledge/request-log.md: date, requester, the request, the decision it serves, type, size, routed-to, and status. This is the record that lets anyone answer "what does the BI team actually spend its time on" three months from now — and the recurring-request pattern in it is the best guide there is to what should become a dashboard or a mart.

Then hand off with the framing intact: the specialist should receive the decision, the deadline, and what's been ruled out — not just the original words. Restating the request is the cheapest quality step in the whole workflow.

Reviewing the log

Every month or so, read knowledge/request-log.md as a dataset rather than a list:

  • Which questions recur? Those are marts and dashboards waiting to be built.
  • Which requesters dominate, and is that the right allocation against the company's priorities?
  • What did the team promise and miss, and why? A pattern of misses is a sizing problem, not an effort problem.
  • What was delivered and never used? Worth knowing before building more of it.

Created by Colin Beck — https://www.linkedin.com/in/beckcolin/

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.