agentleFS
KemetraKemetra/Seshat-BI/.claude/skills/pbip-xray/SKILL.md

pbip-xray

Skill2 starsChanged 4 months agoApache-2.0View on GitHub

Audit a committed PBIP semantic model as text (unused fields, relationship risks, measure-graph findings) and explain a TMDL change in business terms for PR review. Read-only: it parses committed TMDL/PBIR and git history, opens no database, never launches Power BI, and edits nothing.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is 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.

  • 1

    Very specific

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

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

    Absence of evidence never becomes a finding: No parseable report JSON -> X1 wording downgrades to (no report scanned -- visual usage …

    line 56
  2. Distinctive:

    Never edits TMDL/PBIR, never auto-fixes (findings carry fix hints only).

    line 66
  3. Distinctive:

    Never opens a DB connection or the Power BI execution adapter (F016).

    line 67
  4. Distinctive:

    Never grants an approval or moves a readiness stage.

    line 68

≈ “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

pbip-xray

  • Capability id: pbip-xray (docs/capabilities/capabilities.yaml).
  • Authority: advisory. Findings NEVER block, never move a readiness stage, and never carry a numeric health/quality score (hard principle: never fabricate a confidence score). Counts and findings only.
  • Design spec: docs/superpowers/specs/2026-08-01-pbip-xray-design.md.

When to invoke

  • "Audit the Power BI model" / "what is dead weight in this model?"
  • "What changed in the semantic model in this PR, in business terms?"
  • "Is this column used anywhere?" / "why does this measure exist?" (pair with cross-table-lineage for the cross-artifact chain).
  • Reviewing any PR that touches *.SemanticModel/definition/**.tmdl.

The two verbs

seshat xray --format json                 # full-model audit
seshat model-diff --base origin/main --format json   # PR-scoped semantic diff

Exit 0 = ran to completion (however many findings). Exit 3 = could not run; the payload carries {code, message, recovery} blockers (XR001 no committed model; XR002 unresolvable base ref).

Finding families (audit)

idmeaning
X0a table file the parser could not read (excluded, never guessed)
X1column/measure unreferenced in scanned surfaces
X2relationship risks: many-to-many, dead inactive, string keys, snowflake chains
X3measure-graph: circular references, depth >= 5, cross-table duplicate logic
X4hygiene: unmarked date table, default summarizeBy feeding nothing

Diff buckets (model-diff)

semantic (measure logic, column type, relationship behavior), cosmetic (format/description/folder/sort), additive, removed. Relationships match by ENDPOINTS, so TMDL GUID-name churn never reports. Formatting-only DAX churn classifies cosmetic (normalized-body comparison).

The conservative core (read before trusting a finding)

  • Absence of evidence never becomes a finding. No parseable report JSON -> X1 wording downgrades to (no report scanned -- visual usage unknown) and severity drops to info.
  • Unresolved DAX tokens EXCLUDE a column from every "unused" determination.
  • Enforcement stays with seshat check: D6 owns bi-directional relationships (X-Ray skips them entirely), D3 owns duplicate bodies, D7 owns the date-table marker -- X-Ray findings cite them.

What this skill never does

  • Never edits TMDL/PBIR, never auto-fixes (findings carry fix hints only).
  • Never opens a DB connection or the Power BI execution adapter (F016).
  • Never grants an approval or moves a readiness stage.

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.