Tier 4

fmtsb - Formalize Make This Skill Because

FMTSB - Formalize Make This Skill Because

Input: $ARGUMENTS


Core Principles

  1. Drafts are hypotheses, not skills. A skill draft from /mts is a guess about what the skill should be. Formalization tests that guess against quality standards, existing library topology, and usability requirements.

  2. Schema enforcement is quality enforcement. The schema isn’t bureaucracy — each required element exists because skills without it fail in predictable ways. Missing failure modes = blind spots. Missing integration = orphan skill. Missing depth scaling = inconsistent output quality.

  3. Overlap is the default. With 367 skills, most new skills overlap something. The question isn’t “does this overlap?” but “is the overlap acceptable?” Acceptable overlap: different angle on same problem. Unacceptable: same angle, same problem, different name.

  4. Adoption requires discoverability. A skill nobody can find is a skill that doesn’t exist. Formalization must verify naming, categorization, and integration paths that make the skill findable.

  5. Dependencies are bidirectional. What does this skill need from others? What do others need from this skill? Both directions must be mapped.


Phase 1: Draft Assessment

Receive or generate the draft skill (from /mts or direct input).

[F1] DRAFT_SOURCE: [/mts output | direct input | existing skill being upgraded]
[F2] DRAFT_STATUS: [skeleton | partial | near-complete]
[F3] SKILL_ID: [proposed id]
[F4] SKILL_NAME: [proposed full name]
[F5] STATED_PURPOSE: [what the draft says it does]

Phase 2: Schema Enforcement

Check the draft against the required schema. Every element must be present and substantive (not placeholder).

ElementStatusIssue
Frontmatter (name, description)[PASS/FAIL/PARTIAL][if not PASS, what’s wrong]
Core Principles (3-6, specific)[PASS/FAIL/PARTIAL]
Phase structure (numbered findings)[PASS/FAIL/PARTIAL]
Finding prefix (unique letter)[PASS/FAIL/PARTIAL]
Output format (structured)[PASS/FAIL/PARTIAL]
Failure modes table (≥4 entries)[PASS/FAIL/PARTIAL]
Depth scaling table (1x-8x)[PASS/FAIL/PARTIAL]
Pre-completion checklist (≥6 items)[PASS/FAIL/PARTIAL]
Integration section[PASS/FAIL/PARTIAL]
[F6] SCHEMA_SCORE: [N of 9 elements PASS]
[F7] CRITICAL_GAPS: [list of FAIL elements]
[F8] QUALITY_GAPS: [list of PARTIAL elements]

Quality Checks (beyond schema)

[F9] PRINCIPLES_CHECK: Are principles specific to this skill (not generic advice)? [YES/NO]
[F10] FAILURE_MODES_CHECK: Are failure modes based on real failure patterns (not hypothetical)? [YES/NO]
[F11] CHECKLIST_CHECK: Is every checklist item binary-checkable? [YES/NO]
[F12] DEPTH_CHECK: Does depth scaling specify measurable minimums? [YES/NO]

Phase 3: Overlap Detection

Check the proposed skill against the existing library.

[F13] OVERLAP ANALYSIS:
| Existing Skill | Overlap Level | What Overlaps | What's Different |
|---------------|--------------|---------------|-----------------|
| /[id] | [none/low/medium/high/duplicate] | [area] | [area] |
| /[id] | [level] | [area] | [area] |

Overlap Verdicts

  • DUPLICATE: Same problem, same angle → REJECT: use or update existing skill
  • HIGH: Same problem, different angle → JUSTIFY: explain why both should exist
  • MEDIUM: Adjacent problem, some shared territory → ACCEPTABLE: note boundary
  • LOW/NONE: Different problem → PROCEED
[F14] OVERLAP_VERDICT: [proceed | justify | reject]
[F15] OVERLAP_RATIONALE: [why this skill deserves to exist alongside overlapping ones]

Phase 4: Dependency Mapping

[F16] INBOUND DEPENDENCIES (what feeds INTO this skill):
  - /[id] → this skill: [what it provides]
  - /[id] → this skill: [what it provides]

[F17] OUTBOUND DEPENDENCIES (what this skill feeds INTO):
  - this skill → /[id]: [what it provides]
  - this skill → /[id]: [what it provides]

[F18] COMPLEMENTARY SKILLS (good to use alongside):
  - /[id]: [why they complement each other]

[F19] FINDING_PREFIX_COLLISION: [does the prefix collide with existing skills? YES/NO]

Phase 5: Rollout Readiness

[F20] DISCOVERABILITY:
  - Guessable name? [YES/NO — if NO, suggest alternatives]
  - In the right category? [which category]
  - Listed in CLAUDE.md tables? [which table it belongs in]
  - Integration pointers from related skills? [which skills should link to this one]

[F21] ADOPTION BARRIERS:
  - [barrier 1] — [how to address]
  - [barrier 2] — [how to address]

[F22] VALIDATION PLAN:
  - Test case 1: [input] → expected [output type]
  - Test case 2: [input] → expected [output type]

Phase 6: Output

FORMALIZATION REPORT
====================

SKILL: /[id] — [name]
DRAFT_SOURCE: [source]

SCHEMA COMPLIANCE: [N/9] elements pass
CRITICAL_GAPS: [list or "none"]
QUALITY_GAPS: [list or "none"]

OVERLAP_VERDICT: [proceed | justify | reject]
OVERLAP_DETAIL: [key findings]

DEPENDENCIES:
  INBOUND: [list]
  OUTBOUND: [list]

ROLLOUT_READINESS: [ready | needs work]
BLOCKERS: [list or "none"]

FORMALIZED_SKILL: [complete SKILL.md content if ready, or gap list if not]

STATUS: READY_FOR_DEPLOY | NEEDS_REVISION
REVISION_ITEMS:
- [item 1]
- [item 2]

Failure Modes

FailureSignalFix
Rubber stampEverything passes without scrutinySchema check must be element-by-element with evidence
Overlap ignoredNo overlap analysis performedCheck at least 5 potentially overlapping skills
Generic principles accepted”Be thorough” passes as a principleEach principle must be specific to this skill’s domain
Placeholder elements”TBD” or “TODO” in schema elementsEvery element must be substantive, not placeholder
Missing dependenciesSkill exists in isolationFind at least 2 inbound and 2 outbound connections
Name not testedID accepted without guessability checkAsk: would someone looking for this skill try this name?

Depth Scaling

DepthSchema Elements CheckedOverlap Skills ComparedDependencies MappedTest Cases
1x5321
2x9 (all)542
4x9 + quality checks863
8x9 + quality + style12All identified5

Default: 2x. These are floors.


Pre-Completion Checklist

  • Draft assessed with source identified
  • All 9 schema elements checked individually
  • Quality checks performed (principles specific, failure modes real, checklist binary)
  • Overlap analysis against at least 5 candidates
  • Overlap verdict justified
  • Inbound and outbound dependencies mapped
  • Finding prefix checked for collisions
  • Discoverability assessed (naming, categorization, integration links)
  • Validation test cases defined
  • Status is binary: READY_FOR_DEPLOY or NEEDS_REVISION with specific items

Integration

  • Use after: /mts (to formalize a draft)
  • Use after: /cs CREATE (to quality-check a created skill)
  • Differs from /mts: mts creates from scratch; fmtsb formalizes an existing draft
  • Differs from /pci: pci improves many skills systematically; fmtsb formalizes one skill thoroughly
  • Routes to: deployment (save to skills directory) or back to revision