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).
In this repository60 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- approval-consoleSkill
- approval-evidence-packSkill
- business-knowledge-interviewSkill
- capabilitiesSkill
- consumer-data-dictionarySkill
- cross-table-lineageSkill
- dagster-orchestration-adapterSkill
- dashboard-designSkill
- dashboard-intelligenceSkill
- dbt-transformation-adapterSkill
- evidence-pack-generatorSkill
- first-hour-compassSkill
- friendly-pr-reviewerSkill
- grain-confidence-reviewerSkill
- kpi-contract-builderSkill
- pbi-mcp-write-adapterSkill
- pbip-workflowSkill
- pbip-xraySkill
- pbir-authoring-adapterSkill
- portfolio-watchSkill
- powerbi-dashboard-designSkill
- pr-readiness-reviewerSkill
- readiness-viewerSkill
- release-notes-generatorSkill
- report-intent-interviewSkill
- retail-build-warehouseSkill
- retail-discover-portfolioSkill
- retail-governSkill
- retail-initSkill
- retail-onboard-tableSkill
- retail-orchestrateSkill
- retail-scaffoldSkill
- retail-semantic-checkSkill
- retail-validateSkill
- run-next-readinessSkill
- seshat-studioSkill
- showcase-buildSkill
- source-mappingSkill
- speckit-analyzeSkill
- speckit-checklistSkill
- speckit-clarifySkill
- speckit-constitutionSkill
- speckit-git-commitSkill
- speckit-git-featureSkill
- speckit-git-initializeSkill
- speckit-git-remoteSkill
- speckit-git-validateSkill
- speckit-implementSkill
- speckit-planSkill
- speckit-specifySkill
- speckit-tasksSkill
- speckit-taskstoissuesSkill
- bi-analyst-knowledgeSkill
- bi-bigdata-knowledgeSkill
- bi-dax-knowledgeSkill
- bi-python-knowledgeSkill
- bi-sql-knowledgeSkill
- retail-kpi-knowledgeSkill
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
- Distinctive: line 33
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 …
- Distinctive: line 39
Generic: No worked-example specifics (billing codes, segments, PII column names, per-table grain keys).
- Distinctive: line 45
Every cell traces back to an existing committed source -- a JOIN, not a measurement:
- Very specific: line 56
If a per-table source is missing, record that table as not_started / "no evidence yet" -- do NOT invent a status or a number.
- Distinctive: line 84
If asked for "a health score per table" or "one confidence number", DECLINE: cite readiness-model "No fake confidence" and return the four …
- Very specific: line 123
When tables are driven end-to-end, the retail-orchestrate conductor may invoke the control room as the portfolio-level READ after …
≈ “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.
Files
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 checkWARNs,retail validateERRORs (V-RC2/V-RC15/V-RC16), and the per-tabledata-issues.md/blocking-reasons.md/readiness-status.yamlrows. 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 column | Source it aggregates | Severity |
| static WARNs per table | recorded seshat check exit + WARN lines | warning |
| live findings per table | recorded retail validate V-RC2/V-RC15/V-RC16 ERRORs | error |
| data-quality issues | the table's data-issues.md rows | as recorded |
| open blockers | the table's blocking-reasons.md "Open blockers" rows | blocking |
| stage + status | the table's readiness-status.yaml (current_stage, per-stage status) | the four statuses |
| next action | the 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)
| Situation | What the control room does |
| Zero tables with evidence | render an empty roll-up + a clear "no tables onboarded yet" note; do not error, do not invent rows |
| A per-table file malformed / partial | show that table "evidence incomplete: <file>"; do not guess the missing counts |
| A gate result predates the table's last change | show 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 owner | show 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-levelevidence[]/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 checkstatic surface, theretail-validate/retail validatelive 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 underdocs/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.
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
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.

