evidence-pack-generator

Compose a 10-section evidence pack from committed readiness artifacts for Power BI tables.

2|Updated Jun 22, 2026
One-click install
npx skills add https://github.com/Kemetra/Seshat-BI --skill evidence-pack-generator-kemetra
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: evidence-pack-generator
Source: https://github.com/Kemetra/Seshat-BI/tree/main/.claude/skills/evidence-pack-generator
Command: npx skills add https://github.com/Kemetra/Seshat-BI --skill evidence-pack-generator-kemetra

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Late-stage readiness evidence for a table is scattered across a dozen committed artifacts, leaving reviewers without a single legible place to see it all before authorizing publish. This Skill assembles those artifacts into one ordered, readable evidence pack where every claim links back to its source. ## Core Features & Use Cases - Fixed 10-section composition: Builds an ordered pack covering source profile, source map, assumptions, metric contracts, validation results, semantic model, dashboard design, handoff pack, known limitations, and release notes. - Honest status recording: Records each section as not_started, blocked, warning, or pass with evidence and blocking reasons, never fabricating content for missing sources. - Publish-ready surfacing: Displays the publish_ready state and recorded approval read from readiness-status.yaml without writing approvals or moving stages. - Use Case: Before a data-owner sign-off, ask the agent to assemble the evidence pack for a table at Dashboard Ready stage; it writes evidence-pack-index.md and evidence-pack-summary.md under mappings/<table>/ with all blockers surfaced. ## Quick Start Ask the agent to assemble the evidence pack for a specific table so you can review the full late-stage evidence before the data-owner sign-off.

Frequently Asked Questions about evidence-pack-generator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I assemble an evidence pack for a table before publish sign-off?▼

Ask the agent to assemble the evidence pack for the table. It reads readiness-status.yaml and the committed section sources, then writes evidence-pack-index.md and evidence-pack-summary.md under mappings/<table>/ with per-section status and blockers.

What sections does the evidence pack contain?▼

The pack has a fixed ordered contract of 10 sections: source profile, source map summary, assumptions and decisions, metric contracts, validation summary, semantic model summary, dashboard summary, handoff pack, known limitations, and release notes.

Can the evidence pack generator approve a table for publish?▼

No. It only surfaces the publish_ready status and recorded approval read from readiness-status.yaml. It never writes approvals, edits approvals[], or moves any readiness stage to pass; that decision belongs to the named human.

What happens when a section source is missing or unfilled?▼

The section is recorded as blocked with a blocking_reasons entry naming the missing source. The generator never fabricates substitute content, and a blank template is treated the same as a missing source.

Does the evidence pack generator read live databases or Power BI models?▼

No. It reads only committed artifacts in the repository. It never connects to a database, reads a PBIP model, calls the Power BI execution adapter, or publishes anything.

Can I get a numeric readiness or confidence score from the pack?▼

No. The pack reports only the four explicit statuses (not_started, blocked, warning, pass) plus evidence and blocking reasons. Numeric confidence scores and completeness counts are explicitly refused.