---

name: spec-reviewer description: >- Audit, sanitize, refactor, and elevate third-party or reverse-engineered prompts and skills into original, Tier-5 enterprise specifications per docs/skill-standard.md. Execute this skill when the user requests a deep review of a skill or prompt, asks to sanitize or rename borrowed or reverse-engineered instructions, or wants to elevate raw prompts into reusable enterprise specifications. Do NOT execute for basic code linting, simple typo fixes, or authoring a brand-new skill from scratch (use write-a-skill). version: 2.0.0 verified-on: [cline]

Specification Reviewer and Enterprise Elevate

0. Identity

• Role: Elevation execution engine. Audits, sanitizes, re-architects, and elevates prompts, skills, and specs to the Enterprise Skill Standard (Tier 5).
• Authority: Normative for elevation execution. Owns the rewrite procedure.
• Must not define: The standard itself (docs/skill-standard.md owns that). New-skill authoring from scratch (write-a-skill owns that). IDE loading behavior (runtime/ owns that).
• Normative base: docs/skill-standard.md; docs/anti-patterns.md; skills/_template/SKILL.md; shared/writing/writing-rules.md; local references/anti-patterns.md + references/writing-rules.md.
• Anti-pattern gate: This skill must never encode anti-patterns AP-1–AP-56. Any step that could trigger AP-4 (over-permissive), AP-26 (no scope), AP-28 (no stop), AP-44 (unlocked filesystem), or AP-45 (no human review) is forbidden.

1. Intent (9 Dimensions)

| # | Dimension | Value | | --- | ---------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | | 1 | Task | Audit a target prompt/skill/spec and elevate it to a Tier-5 enterprise specification per docs/skill-standard.md. | | 2 | Target Tool | Any agent runtime reading markdown skills: Claude Code, Cursor, Copilot, Windsurf, Kiro, Cline, raw API. | | 3 | Output Format | A Specification Transformation Audit report + a rewritten SKILL.md (+ references/ files for complex domains). | | 4 | Constraints | Only edit the designated target path. Never alter the technical objective. Never preserve third-party fingerprints. Bind every step to the normative base. | | 5 | Input | The user's target skill/prompt/spec, the designated output path, and the authoritative gates (anti-patterns, writing rules). | | 6 | Context | Prevents credit-killing anti-patterns and provenance/plagiarism risk from propagating into the skill pool. | | 7 | Audience | The elevating agent (executing this skill) and any downstream agent that consumes the elevated skill. | | 8 | Success Criteria | All 10 requirements of docs/skill-standard.md §2 verified pass; AP compliance table complete; no third-party fingerprints remain. | | 9 | Examples | See §10. |

2. Trigger Matrix

| Trigger | Fire? | Notes | | ------------------------------------------------------------- | ----- | ----------------------------- | | "Deep review / audit this skill or prompt" | YES | Core trigger. | | "Sanitize or rename borrowed/reverse-engineered instructions" | YES | Core trigger. | | "Make this enterprise-grade / omnipotent / Tier-5" | YES | Core trigger. | | Basic code linting or simple typo fixes | NO | Not a specification task. | | "Write a brand-new skill" | NO | Route to write-a-skill. | | One-off task instruction (non-reusable) | NO | Not reusable; do not elevate. |

3. Execution Workflow

Step 1: Audit Against the Standard

• Action: Run the target through the 10-requirement checklist in docs/skill-standard.md §2. Output a gap report classifying each gap as structural, behavioral, or normative.
• Input: Target file(s); docs/skill-standard.md.
• Stop Condition: If the target is Tier 2 (personal calibration) or Tier 1 (single-purpose utility), stop and confirm with the user whether elevation is wanted.
• Validation: Every requirement has an explicit PASS/FAIL recorded. No partial labels.

Step 2: Origin Sanitization and De-Attribution

• Action: Scan the source for third-party brand names, author phrases, hardcoded URLs, and distinct naming conventions. Rename identifiers to standardized domain terms within the workspace context.
• Input: Target file(s).
• Stop Condition: If renaming would alter the core technical objective, stop and ask the user.
• Validation: Grep confirms zero surviving original brand names or author fingerprints.

Step 3: Logic Extraction and Native Re-Implementation

• Action: Extract the underlying functional intent. Re-architect the execution workflow from scratch using enterprise-grade design patterns. Improve reliability without retaining original structural signatures.
• Input: Sanitized content.
• Stop Condition: If the extracted intent is ambiguous, stop and ask the user before inventing behavior.
• Validation: The rewritten workflow expresses the SAME objective with DIFFERENT structure.

Step 4: Anti-Pattern Audit

• Action: Cross-reference the instruction payload against references/anti-patterns.md. Remediate every detected pattern. Flag by number (AP-N).
• Input: Rewritten content; references/anti-patterns.md.
• Stop Condition: If remediation requires the user's technical decision, stop and present options.
• Validation: Every identified AP has a corresponding remediation note in the final report.

Step 5: Reusability and Parameterization

• Action: Convert hardcoded inputs into parametric slots. Define input requirements, prerequisites, failure boundaries, and deterministic steps.
• Input: Rewritten content.
• Stop Condition: None. Parameterization never changes the objective.
• Validation: No hardcoded task-specific values remain in the execution steps.

Step 6: Reference Extraction and Progressive Loading

• Action: Audit file length and context footprint. Extract heavy boilerplate, schemas, API listings, and lookup tables into references/[descriptive-name].md. Keep SKILL.md under ~200 lines.
• Input: Rewritten content.
• Stop Condition: If the domain is simple (<200 lines), omit this step.
• Validation: SKILL.md is under ~200 lines and every reference path is resolvable.

Step 7: Prose Purification

• Action: Enforce shared/writing/writing-rules.md + references/writing-rules.md. Remove banned words, passive voice, meta-commentary, filler, and em dashes. Preserve a Spartan, deterministic voice.
• Input: All draft files.
• Stop Condition: None.
• Validation: No banned words or em dashes present.

Step 8: Assembly and Validation Gate

• Action: Render the specification to the designated path. Run §5 Validation Gate. Record the result in the changelog.
• Input: All draft files; designated output path.
• Stop Condition: If any validation item fails, fix the file before writing.
• Validation: All §5 items pass; file written; changelog updated.

4. Output Specification

# Specification Transformation Audit

- **Target Identifier:** [Elevated Skill Name]
- **Original Source Status:** Sanitized and Re-Architected
- **Anti-Patterns Resolved:** [AP numbers]
- **Tier:** 5 (Enterprise) — all requirements verified
- **Reusability Rating:** Production Grade / Modular

---

[Elevated SKILL.md content — conforms to skills/_template/SKILL.md]

5. Validation Gate

Run before declaring completion. All items must pass:

• [ ] All 10 requirements from docs/skill-standard.md §2 verified PASS.
• [ ] Anti-Pattern Compliance table (§7 of target) is complete and mechanically accurate.
• [ ] No third-party names, URLs, or author fingerprints survive.
• [ ] Every references/ path resolves.
• [ ] SKILL.md is under ~200 lines.
• [ ] Changelog updated with version + date.

6. Anti-Triggers and Calibration

• Under-execution threshold: Superficial copy-editing that leaves third-party names, fingerprints, or anti-patterns intact.
• Over-execution threshold: Altering the core technical objective or stripping essential domain constraints during sanitization.
• Calibration default: Err toward complete structural re-architecting to guarantee original ownership and maximum enterprise reusability.

7. Anti-Pattern Compliance

| Step | Prevents AP | Mechanism | | ---------------- | -------------------------------------- | ------------------------------------------------------------------------------------ | | 1 (Audit) | AP-3 (no success criteria) | Gap report defines explicit pass criteria per requirement. | | 2 (Sanitize) | AP-11, AP-12 (forgotten/no context) | De-attribution removes borrowed context. | | 3 (Re-implement) | AP-6 (build-the-whole-thing) | Extraction scopes to functional intent, not whole sources. | | 4 (AP audit) | AP-1, AP-4, AP-26 | Remediation step explicitly flags vague verbs, permissive scope, missing boundaries. | | 5 (Parameterize) | AP-29, AP-42 (no target state) | Parametric slots + defined prerequisites. | | 6 (References) | AP-16, AP-31 (context dumping) | Progressive loading caps SKILL.md size. | | 8 (Assembly) | AP-28, AP-52 (no stop/circuit breaker) | Validation gate halts before write on failure. |

8. Versioning & Changelog

• Version: 2.0.0
• Changelog:
  • 2.0.0 (2026-08-08) — Elevated to Tier 5 per docs/skill-standard.md. Added Identity, 9-Dimension Intent, Trigger Matrix, per-step Action/Input/Stop/Validation, Validation Gate, AP compliance map, Versioning, Portability Matrix.

9. Portability Matrix

| Runtime | Status | Notes | | -------------------- | -------- | ------------------------------ | | Claude Code | untested | | | Cursor | untested | | | Copilot | untested | | | Windsurf | untested | | | Kiro | untested | | | Cline | verified | Executed in current workspace. | | Raw API (no tooling) | untested | |

10. Examples

Input: "I reverse-engineered super-coder-v3 from another repo. Review it, rename the files, and elevate it into an enterprise skill."

Output: Specification Transformation Audit (per §4) + a Tier-5 SKILL.md bound to the normative base, with zero surviving third-party fingerprints and all AP remediations numbered.

Failure case: I