| name | scorecard |
| description | 8-layer readiness audit for Kapi platform. Audits actual code paths (manifest-transformer → runtime) and rewrites scorecard.md with honest percentages. Use when status.md grades feel optimistic, before /prd, or after significant code changes. |
| allowed-tools | Read, Grep, Glob, Write |
/scorecard — 8-Layer Readiness Audit
Update docs/operations/scorecard.md by auditing the actual code paths for all 8 layers.
What You're Measuring
Three columns, each as an honest percentage:
| Column | Question |
|---|
| Manifest → Config | Does lib/manifest-transformer.ts extract this layer's data and put it into TenantConfig? |
| Config → Runtime | Does the runtime code (chat route, tenant page, graph router) actually read and use that config? |
| End-to-End | Both columns working together. Not an average — if either side is broken, E2E is low. |
Scoring guidance:
- 0%: Not extracted / not read at all
- 25%: Partially extracted or read but not acted on
- 50%: Extracted and read, but with significant gaps or hardcoded fallbacks
- 75%: Works end-to-end for common cases, some edge cases missing
- 95%: Fully wired, only minor gaps
Audit Protocol
For each layer (1-8), read the corresponding file at layers/{NN}-{name}.md. It lists the files to read and the key question to answer. Assess both sides of the wire independently, then write a one-sentence verdict that explains the biggest remaining gap.
Output
After auditing all 8 layers, write the updated scorecard using the format in templates/output.md.
Be honest. If something is 0%, say 0%. The scorecard is used to prioritize sprint work — optimism defeats its purpose.
Blackboard Auto-Post
Read ../blackboard/posting.md for the posting convention. After completing the audit:
- If any layer's E2E score dropped >20% from the previous scorecard: pause and ask the human via AskUserQuestion before continuing. Then post as
finding.
- If the Top 3 Blockers changed significantly: auto-post as
finding.
- If scores are stable: do NOT post — no news is good news.