agentleFS
KemetraKemetra/Seshat-BI/.claude/skills/retail-control-room/SKILL.md

retail-control-room

Skill2 starsChanged 4 months agoApache-2.0View on GitHub

Show the consolidated, cross-table data-quality control room for the Seshat BI repo -- one worst-first roll-up of every table's stage, status, static WARNs, live findings, open blockers, and next action, plus a portfolio-wide open-blockers list. Use when someone asks "show the control room", "what is broken across all tables", or "which table do I fix next". READ-ONLY and invoke-and-present only: it AGGREGATES evidence that already exists (per-table readiness-status.yaml / data-issues.md / blocking-reasons.md + recorded seshat check / retail validate results); it runs NO new validator, edits NO per-table file, clears NO blocker, and emits NO fabricated health/confidence score (every cell is a measured number traceable to a committed source).

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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:

    No fake confidence: Every line carries a MEASURED NUMBER as evidence (a row count, an orphan count, a penny delta, an open-blocker count, a …

    line 33
  2. Distinctive:

    Generic: No worked-example specifics (billing codes, segments, PII column names, per-table grain keys).

    line 39
  3. Distinctive:

    Every cell traces back to an existing committed source -- a JOIN, not a measurement:

    line 45
  4. Very specific:

    If a per-table source is missing, record that table as not_started / "no evidence yet" -- do NOT invent a status or a number.

    line 56
  5. Distinctive:

    If asked for "a health score per table" or "one confidence number", DECLINE: cite readiness-model "No fake confidence" and return the four …

    line 84
  6. Very specific:

    When tables are driven end-to-end, the retail-orchestrate conductor may invoke the control room as the portfolio-level READ after …

    line 123

≈ “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

retail-control-room

The per-table scorecard answers "where is THIS table"; the control room answers "where is the WHOLE portfolio, and which table do I touch next". It is a read-only JOIN over evidence that already exists per table -- the portfolio-level sibling of readiness-scorecard.md. It introduces no new validator and no new gate (roadmap rule 8; Principle VIII); it aggregates and presents, then STOPS.

Scope boundary (read first)

  • Aggregates, never re-derives. The only evidence shown is evidence already committed: seshat check WARNs, retail validate ERRORs (V-RC2/V-RC15/V-RC16), and the per-table data-issues.md / blocking-reasons.md / readiness-status.yaml rows. It runs NO new check.
  • Read-only. It never edits a per-table artifact, never clears a blocker, never writes a pass, never runs SQL, never opens a DB connection. Clearing a blocker stays a per-table action by its named owner (Principle V).
  • No fake confidence. Every line carries a MEASURED NUMBER as evidence (a row count, an orphan count, a penny delta, an open-blocker count, a finding-id) -- never an adjective, never a fabricated health score (roadmap rule #9). A numeric score is OPTIONAL and DEFERRED; the control room MUST NOT emit one.
  • Every cell traces to a committed source. A number with no traceable source path (and row/line where applicable) is a defect.
  • Generic. No worked-example specifics (billing codes, segments, PII column names, per-table grain keys). C086 is a filled instance cited as a reference, never baked in.
  • ASCII only, UTF-8 no BOM.

Aggregates, never re-derives (the evidence chain)

Every cell traces back to an existing committed source -- a JOIN, not a measurement:

Control-room columnSource it aggregatesSeverity
static WARNs per tablerecorded seshat check exit + WARN lineswarning
live findings per tablerecorded retail validate V-RC2/V-RC15/V-RC16 ERRORserror
data-quality issuesthe table's data-issues.md rowsas recorded
open blockersthe table's blocking-reasons.md "Open blockers" rowsblocking
stage + statusthe table's readiness-status.yaml (current_stage, per-stage status)the four statuses
next actionthe table's readiness-status.yaml next_action--

If a per-table source is missing, record that table as not_started / "no evidence yet" -- do NOT invent a status or a number. The control room cannot UPGRADE a status; a pass it shows must be backed by the same evidence the per-table file carries.

Run it -- produce the consolidated view

Render templates/data-quality-control-room.md filled from the per-table evidence.

1. Per-table roll-up (one row per table, worst-first)

Scan each mappings/<table>/. For each table emit a row: table id, source family, current stage, stage status, count of static WARNs, count of live findings, count of open blockers, single next action. Every count is a MEASURED number copied from the source (e.g. open-blocker count = the number of "Open blockers" rows in that table's blocking-reasons.md; never "several"). Sort blocked above warning above pass.

2. Portfolio open-blockers list (the action half)

List EVERY open blocker across all tables with {table, stage blocked, concrete reason, measured evidence, named owner} copied from the source blocking-reasons.md, ordered worst-first: an error-level live finding (V-RC2/V-RC15/V-RC16) outranks a warning static WARN (proven defect > suspect pattern). Never self-assign an owner; a blocker with no owner is shown "UNASSIGNED" and flagged.

3. Traceability

For every cell, be able to name the exact committed source path (and row/line). If asked "where does this number come from", answer with the path -- not a recomputation.

No fake confidence (the guardrail)

If asked for "a health score per table" or "one confidence number", DECLINE: cite readiness-model "No fake confidence" and return the four explicit statuses + the measured counts with their source paths instead. A roll-up is exactly where a tidy invented score is tempting; it is forbidden.

Honest-state rules (never invent, never silently re-run)

SituationWhat the control room does
Zero tables with evidencerender an empty roll-up + a clear "no tables onboarded yet" note; do not error, do not invent rows
A per-table file malformed / partialshow that table "evidence incomplete: <file>"; do not guess the missing counts
A gate result predates the table's last changeshow the recorded result + its timestamp; mark live findings "not run since <date>"; do NOT run the live check itself (Principle VIII -- the human's call)
Two sources disagree (status pass vs an open blocker)SURFACE the conflict as a finding; do NOT resolve it by picking one (Principle V)
A blocker with no named ownershow owner "UNASSIGNED" and flag it; never self-assign

Read-only proof

The skill modifies nothing: after a run, git status shows zero modified per-table mappings/<table>/, data-issues.md, or blocking-reasons.md files. It triggers no state-mutating seshat check / retail validate run of its own.

See also

  • The output shape: templates/data-quality-control-room.md; the per-table sibling: templates/readiness-scorecard.md.
  • The aggregated sources: templates/data-issues.md, templates/blocking-reasons.md, templates/readiness-status.yaml (top-level evidence[] / blocking_reasons[]).
  • The model + no-fake-confidence rule: docs/readiness/readiness-model.md; the stage sequence: docs/readiness/readiness-pipeline.md.
  • The gates it reads (never re-runs as a new check): the retail-govern / seshat check static surface, the retail-validate / retail validate live surface (src/seshat/validate.py, V-RC2/V-RC15/V-RC16).
  • The conductor it plugs into: .claude/skills/retail-orchestrate/SKILL.md.
  • The roadmap row + hard rules: docs/roadmap/roadmap.md (F012, Layer 4; #7/#8/#9); Principles V, VII, VIII. A filled worked example lives under docs/worked-examples/.

Orchestration

When tables are driven end-to-end, the retail-orchestrate conductor may invoke the control room as the portfolio-level READ after sequencing a table -- to see the whole board and the next blocker to clear. This skill reads state and reports; it advances no stage and clears no blocker. The self-heal loop and any per-table fix live in retail-orchestrate / the per-table owner, never here.

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.