| name | flow-pack-agent-team-scoring |
| description | FPAT scoring rubrics — the 7-dimension execution rubric for shaping subtasks inside an epic, and the 5-dimension continuation rubric (ship gate >= 40) for deciding which epics ship. Use when running /fpat-plan-issue, /fpat-continuation, or otherwise scoring FPAT directions, subtasks, or epics. No Python — the agent applies the rubric inline. |
| metadata | {"author":"w7-mgfcode","version":"1.0.0","scope":"project-local"} |
Flow-Pack Agent-Team Scoring
Project-local skill. Hand-authored for the CVForge AI FPAT workflow — not installed via
npx skills add, so it is intentionally absent from skills-lock.json. Do not let
npx skills update manage or overwrite it. Scaffolded with npx skills init per
.claude/rules/skills.md ("Authoring a project-local skill").
When to use
- Scoring candidate directions or subtasks inside a chosen epic → use the
7-dimension execution rubric.
- Deciding which epics ship in a V1→V2 continuation pass → use the
5-dimension continuation rubric and its ship bands.
- Invoked by
/fpat-plan-issue (step 4) and /fpat-continuation.
When NOT to use
- Do not write Score values to the board from here. Board field writes are out of scope; the
board's
Score field is set deliberately, and fpat-project-sync.yml never touches Status/Score
(see .claude/docs/flow-pack-agent-team/board-spec.md).
- Do not invent a Python scorer. Scoring is applied inline by the agent.
7-Dimension Execution Rubric (subtask / direction shaping)
Score each dimension 1–10. Higher = more favorable (frame Risk and Dependency load as
low-burden favorability, so low risk → high score). Source:
.claude/docs/flow-pack-agent-team/execution-pipeline.md.
- Value — delivered value of the direction.
- Risk — favorability of the risk profile (low risk scores high).
- Readiness — how ready the work is to start (deps met, scope clear).
- Dependency load — favorability of the dependency burden (few deps score high).
- Validation ease — how cheaply the result can be validated.
- Rollback safety — how safely the change can be reverted.
- Evidence strength — how well claims are backed by file paths / command output.
Use this rubric to compare directions and to sanity-check that each subtask earns its place.
5-Dimension Continuation Rubric (epic ship gate)
Score each dimension 1–10. Source:
.claude/docs/flow-pack-agent-team/continuation-discipline.md.
- Value
- Risk
- Readiness
- Complexity
- Evidence
Ship bands (keep verbatim for the first CVForge AI adoption umbrella)
>= 40 — ship.< 36 — defer.36–39 — negotiate with the user.
The board's Score field stores this 5-dimension total; fpat-blocked-sweep.yml audits epics
that left Backlog with Score < 40.
Single source of truth for the bands. These three bands are the verbatim set that
/fpat-continuation applies at its step 7 (see .claude/commands/fpat-continuation.md). Keep
them byte-identical across this skill, that command, and
.claude/docs/flow-pack-agent-team/continuation-discipline.md — if one changes, change all three.
Instructions
- Pick the rubric: 7-dimension for subtasks/directions, 5-dimension for epic ship decisions.
- Score every dimension 1–10 with a one-line justification grounded in evidence.
- Sum the total; for the 5-dimension rubric apply the ship bands above.
- Present the scores as a table; never write them to GitHub or the board from this skill.
