agentleFS
KemetraKemetra/Seshat-BI/.claude/skills/report-intent-interview/SKILL.md

report-intent-interview

Skill2 starsChanged 4 months agoApache-2.0View on GitHub

Given approved metric contracts and a ready semantic model, interview the report requester to capture what a dashboard report is FOR -- audience, purpose, supported decision, review cadence, primary business questions, outcome/driver/guardrail metrics (referenced by name only, never defined), comparisons, dimensions/filters, expected actions/exceptions, pages/drill paths, and mobile/accessibility/language/RTL needs -- and record the result as a committed Report Intent artifact plus a named-human `report_intent_approval` decision. Use when someone describes a report they need in plain language, before any page blueprint, visual spec, or report composition is authored. It asks obvious low-risk items as one batch and critical items individually with explicit named-human approval; it masks PII by default; it records the intent and STOPS. It does NOT define metric meaning, does NOT self-grant any approval, and does NOT emit a confidence score.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 3 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Before dashboard-design, the dashboard coordinator, page blueprints, visual specs, or report composition -- those stages trace their …

    line 40
  2. Very specific:

    Reference metrics by name only: Every outcome_metrics / driver_metrics / guardrail_metrics entry names an approved metric contract (name + …

    line 76
  3. Very specific:

    Record the artifact and the decision: Write mappings/<subject-area>/design/report-intent.yaml per templates/report-intent.yaml, and record …

    line 86
  4. Very specific:

    Never self-grant report_intent_approval (Principle V.

    line 104
  5. Distinctive:

    Never advance a readiness stage.

    line 106
  6. Very specific:

    Never emit a numeric confidence score (never_fabricate_a_confidence_score).

    line 107
  7. Very specific:

    Never define metric meaning here (route to retail-kpi-knowledge / the metric-contract layer).

    line 108
  8. Distinctive:

    Never display raw suspected-PII values by default, and never commit them.

    line 110
  9. Distinctive:

    Never invent a metric contract to make a reference resolve.

    line 111

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

report-intent-interview

The Report Intent Interview is the missing front end of the dashboard intelligence journey (spec 123, US1). It converts a conversational report request into a committed, reviewable report-intent.yaml so every downstream artifact (blueprint, composition, preview, audit, PBIR) traces back to a clear, approved report purpose instead of an unrecorded assumption.

This skill mirrors .claude/skills/business-knowledge-interview/SKILL.md structurally (load-existing-first, batch/critical question grouping, PII masking, record + STOP, never self-grant) but does NOT extend it: the required_inputs and focus differ. Behavior is governed by the contract contracts/interview/report-intent-interview.yaml; the shape of the produced artifact is templates/report-intent.yaml. The static rule DL9 (seshat check) verifies the recorded artifact's SHAPE, not the conversation.

When to use

  • Someone describes a report they need in plain language ("I need a weekly report for branch managers to spot underperforming branches").
  • Before dashboard-design, the dashboard coordinator, page blueprints, visual specs, or report composition -- those stages trace their business_question back to a question declared here (FR-002a) and are blocked until this intent's own report_intent_approval decision is approved.

Preconditions (stop if unmet)

  • Approved metric contracts exist for the subject area (readiness.status: pass). No approved contract for a metric the request needs => record a gap, route upstream to metric-contract definition, and leave the intent blocked on that metric (FR-004) -- never define the metric here.
  • The semantic model is ready (semantic_model_ready: pass). Not ready => stop, name the missing readiness, and do not proceed to design.
  • Load the existing Report Intent (if any) first. An existing intent is presented for confirmation or supersession -- never silently overwritten.

How it runs

  1. Ground every question in the request and the approved contracts. Focus on: audience, purpose, supported decision, review cadence, primary business questions, metric roles by name (outcome/driver/guardrail), comparisons, dimensions and filters, expected actions and exceptions, pages/drill paths, mobile/accessibility/language/RTL needs, and exclusions/non-goals.
  2. Refuse to commit a vague request. "Build an executive dashboard" gets focused disambiguating questions; the intent is NOT committed until audience, purpose, and at least one primary business_questions entry are resolved (US1 AC#4).
  3. Batch the obvious low-risk items into one owner confirmation (e.g. comparison phrasing, filter defaults). Critical decision types are never in a batch. The owner may exclude any item; the remainder is approved in one action and each excluded item becomes an individual pending question.
  4. Ask critical decisions individually. The intent's own approval (report_intent_approval) is a critical decision type, authority class report_owner (contracts/knowledge/approval-authority.yaml); it requires an explicit, per-decision approval from a named human -- the agent never self-grants it (Principle V).
  5. Reference metrics by name only. Every outcome_metrics / driver_metrics / guardrail_metrics entry names an approved metric contract (name + store_ref + status_required: pass) -- the same triple shape as dashboard-page-blueprint.yaml's required_metric_contracts. There is no formula/DAX field. A metric with no approved contract is a gap: record it, route it upstream, and leave the intent blocked on that metric -- never invent the metric here.
  6. Mask suspected PII by default. Any PII-adjacent dimension/filter follows the shipped masking behavior; unmasking requires an explicit owner instruction, recorded as a pii_handling decision.
  7. Record the artifact and the decision. Write mappings/<subject-area>/design/report-intent.yaml per templates/report-intent.yaml, and record the report_intent_approval decision in the project Decision Store once a named human of authority class report_owner approves. Unanswered items stay pending / needs_user_input; a metric gap stays blocked with a matching blocking_reasons entry.

Exit

Report the current gate verdict for the report_intent stage (contracts/report/report-intent.yaml) and the next allowed stage (dashboard_blueprint) -- which stays blocked until this intent's report_intent_approval is valid (an unapproved intent yields blocked, never a false pass).

Hard stops (never)

  • Never self-grant report_intent_approval (Principle V; never_self_grant_approval).
  • Never advance a readiness stage.
  • Never emit a numeric confidence score (never_fabricate_a_confidence_score).
  • Never define metric meaning here (route to retail-kpi-knowledge / the metric-contract layer).
  • Never display raw suspected-PII values by default, and never commit them.
  • Never invent a metric contract to make a reference resolve; record the gap and leave the intent blocked instead (FR-003/FR-004).

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.