name: agentflow-fipd
description: Classify findings using the FIPD taxonomy (Fix/Investigate/Plan/Decide) instead of severity-only ratings. Background knowledge for all analysis and audit skills.
user-invocable: false

FIPD Finding Taxonomy

Every finding gets classified by action type — what the reader should do next, not just how bad it is.

The Four Actions

Rules

1. Every finding gets exactly one FIPD classification — never leave findings unclassified
2. Fix items go directly to TODO-Today — no backlog detour needed
3. Investigate items get a timebox — max 1 hour before escalation
4. Plan items go to BACKLOG#Ideation — they need design work before implementation
5. Decide items surface immediately to user — the agent cannot resolve trade-offs alone

Uncertainty Declaration

All findings must declare what remains unknown or unverified:

[Action]: [finding] -- Unknown: [what remains unverified]

• Unknown: clause is mandatory for Investigate and Decide actions
• Unknown: clause is recommended for Fix and Plan actions
• Omitting uncertainty leads to false confidence and wasted effort

Examples

Fix: Missing input validation on collection name field -- Unknown: none, root cause confirmed
Investigate: Intermittent 500 on batch export -- Unknown: whether it's memory or timeout related
Plan: Need rate limiting on AI compose endpoint -- Unknown: expected request volume at scale
Decide: Should tags be stored denormalized for read speed or normalized for consistency? -- Unknown: actual read/write ratio in production

Integration with Other Skills

• /agentflow-dod requires FIPD classification on all quality findings
• /agentflow-loop cleanup sub-loop uses FIPD to decide fix vs. pause
• Known patterns in KNOWN_PATTERNS.md carry FIPD action tags

See taxonomy.md for extended examples and edge cases.