``` MONNA SKILL AUDIT REPORT™ ══════════════════════════════════════════════════════════════════ Skill audited: monna-course-architect File path: user-provided Framework basis: AGENT SKILL BUILDER FRAMEWORK v1.1 Audit date: June 19, 2026 [MONNA-Analysis-2026] ══════════════════════════════════════════════════════════════════ DIMENSION SCORES ──────────────────────────────────────────────────── DISCOVERABILITY 12/15 [PASS] STRUCTURE 11/15 [WARN] INSTRUCTION CLARITY 13/15 [WARN] MODEL PORTABILITY 9/10 [PASS] VALUE DENSITY 10/10 [PASS] CST COMPLIANCE N/A ──────────────────────────────────────────────────── TOTAL: 55/75 [PROFESSIONAL] TRIGGER ANATOMY (v1.1 Five-Part Check) ──────────────────────────────────────────────────── PART 1 Action Core: ✓ PRESENT PART 2 Signal Tokens: ✓ PRESENT PART 3 Casual Phrasings: ⚠ WEAK (1 present — "I want to teach X"; needs ≥2) PART 4 Indirect-Intent: ✓ PRESENT PART 5 Negative Boundary: ✗ ABSENT — SHIP-BLOCKER SHIP-BLOCKERS ──────────────────────────────────────────────────── BLOCKER-01: PART 5 NEGATIVE BOUNDARY MISSING The description contains no explicit "Do NOT use / different from" clause naming adjacent sibling skill territory. The SCOPE section at the bottom of the body does gesture at exclusions, but that is in the body — not in the description field where the trigger engine reads it. A sibling like monna-prompt-to-skill, monna-seo, or any content-build skill could be mis-selected when a user asks to "build something" or "design a program." This is a hard ship-blocker. Must be resolved before publishing. COLLISION RISKS ──────────────────────────────────────────────────── MODERATE — monna-prompt-to-skill Token overlap: "build", "design", "plan", "structure". A user saying "build me a training program" could ambiguously trigger either skill. Fix: add a Part 5 clause to the description field naming monna-prompt-to-skill (for building AI skills/prompts) as distinct from this skill (for building human-facing learning experiences). LOW — listing-health-auditor / monna-seo Stage 6 PACKAGING section introduces tokens ("Gumroad", "publish", "title", "promise") that overlap with listing and SEO skills. Risk is low because Stage 6 is optional and gated, but worth a brief mention in Part 5. LOW — any content-generation skill The SCOPE section disclaims "does NOT write full course content/scripts" but this exclusion is in the body, not the description. If a sibling content-build skill exists, add it to Part 5. FINDINGS (ordered by priority) ──────────────────────────────────────────────────── FINDING-01 [CRITICAL]: PART 5 NEGATIVE BOUNDARY ABSENT IN DESCRIPTION FIELD Location: description: field Issue: The exclusion language ("It does NOT: write the full course content/scripts… produce marketing copy… invent market research") exists only in the body SCOPE section. The description field — the only layer the trigger engine reads — contains no "Do NOT use this for / Use X instead" clause naming any sibling skill by name. This is the ship-blocker. Fix: Append the following (or equivalent) to the end of the description field, inside the | block: "Do NOT use for writing full course scripts or content modules (use a content-build skill), producing listing copy or marketing pages (use monna-seo or listing skills), or generating live market research (requires a live search tool). Different from monna-prompt-to-skill, which builds AI agent skills, not human learning experiences." FINDING-02 [HIGH]: PART 3 CASUAL PHRASINGS WEAK (1 of ≥2 required) Location: description: field Issue: The description includes "I want to teach X" as an implied casual phrasing but does not include a second distinct informal example. Tired users may phrase requests as "help me put together something people can learn from" or "I have an idea I want to turn into lessons" — neither is covered. Fix: Add at least one more explicit casual phrasing, e.g.: "Also trigger when asked: 'I have this idea, how do I turn it into something teachable?' or 'Help me figure out what to put in my course.'" FINDING-03 [HIGH]: REFERENCED FILE NOT DECLARED IN FRONTMATTER Location: frontmatter + body (CORE FRAMEWORKS section) Issue: The body references `references/course-frameworks.md` in two places (CORE FRAMEWORKS table and Stage 2 instruction) but the frontmatter contains no `references:` or `dependencies:` field declaring it. If the file is absent, the skill silently degrades at Stage 2 with no error path. Fix 1: Add to frontmatter — e.g. `references: references/course-frameworks.md` Fix 2: Add an Error Handling entry: "IF references/course-frameworks.md is not found, proceed using inline Bloom's verb guidance and flag to user that the reference file is missing." FINDING-04 [HIGH]: ERROR HANDLING SECTION ABSENT Location: body Issue: There is no dedicated Error Handling section. The framework requires explicit handling for: missing input (user gives no topic), ambiguous input (topic too broad to scope), and malformed/conflicting constraints (e.g. "teach machine learning to beginners in 30 minutes"). GUARD 1 partially covers one case, but it is not a formal error-handling block. Fix: Add an ## Error Handling section after ## GUARDRAILS with at minimum: — "No topic provided: ask one targeted question before proceeding." — "Topic irreconcilably over-scoped for stated time: state this explicitly, offer two realistic reframes, do not proceed to Stage 2 until resolved." — "Reference file missing: see FINDING-03 fix above." FINDING-05 [MEDIUM]: FRONTMATTER MISSING REQUIRED FIELDS Location: frontmatter (--- block) Issue: The following v1.1 required fields are absent: version, author, license, pricing, compatibility, categories, cst_risk_profile. Only name and description are present. Fix: Add all missing fields. Minimum viable addition: ``` version: 1.0.0 author: [your handle] license: proprietary pricing: [tier] compatibility: claude-code, claude.ai categories: [education, instructional-design, course-creation] cst_risk_profile: none ``` FINDING-06 [MEDIUM]: OUTPUT FORMAT DOES NOT SPECIFY FILE TYPE TRIGGER CONDITION Location: ## OUTPUT FORMAT section Issue: The output format says "For a large course, deliver this as a file (md or docx). For a small one, inline is fine." The threshold for "large" vs "small" is undefined. The model will guess, inconsistently. Fix: State an explicit threshold, e.g.: "If the curriculum map exceeds 4 modules or the user's scope is ≥8 hours of content, deliver as a .md file using create_file. Otherwise deliver inline." FINDING-07 [MEDIUM]: STAGE 4 TOOL NOT NAMED FOR FILE CREATION Location: WORKFLOW → STAGE 4 and OUTPUT FORMAT Issue: The output format instructs delivery as "a file (md or docx)" but no step names the tool (create_file) for doing so. Tool specificity is required per DIMENSION 3. Fix: In the OUTPUT FORMAT section, add: "Deliver using the `create_file` tool, naming the file `[slug]-course-blueprint.md`." FINDING-08 [LOW]: QUALITY GATES SECTION ABSENT Location: body Issue: The framework requires a Quality Gates checklist. None is present. Fix: Add a ## Quality Gates section with checks such as: — [ ] Scope statement produced before Stage 2 — [ ] Every objective uses a Bloom's-compliant measurable verb — [ ] Alignment table produced in Stage 5 — [ ] GUARD 2 respected (no fabricated market claims in output) FINDING-09 [LOW]: EXAMPLES SECTION ABSENT (minimum 2 required) Location: body Issue: DIMENSION 2 requires at least 2 worked examples. None are present. The skill is complex (6 stages) — even abbreviated worked examples (a narrow topic vs a broad domain) would materially reduce mis-triggering and set output quality expectations. Fix: Add an ## Examples section with at minimum two short cases: Example A — narrow topic ("Teach people to use Notion for task management"): show a 3-line scope statement and 3 objectives. Example B — domain-only input ("I want to make a course about marketing"): show how Stage 0 plays out with 3 course patterns presented. SUMMARY ──────────────────────────────────────────────────── Findings total: 9 (CRITICAL: 1 / HIGH: 3 / MEDIUM: 3 / LOW: 2) Ship status: BLOCKED (CRITICAL ship-blocker present; resolve FINDING-01 first) Estimated fix time: 45 min – 1.5 hours — FINDING-01 (Part 5 in description): ~10 min — FINDING-02 (Part 3 casual phrasing): ~5 min — FINDING-03 + 07 (reference declaration + create_file): ~10 min — FINDING-04 (Error Handling section): ~15 min — FINDING-05 (frontmatter fields): ~10 min — FINDING-06 (output threshold): ~5 min — FINDING-08 + 09 (QA gates + examples): ~20 min ══════════════════════════════════════════════════════════════════ ``` --- **Auditor note:** The core skill logic is genuinely strong — the two Honesty Guards, Backward Design sequencing, and Bloom's alignment are all well-specified and would produce measurably better output than a direct prompt. That's why VALUE DENSITY scores 10/10. The gap between PROFESSIONAL and ENTERPRISE-READY is almost entirely structural (missing frontmatter, no Error Handling, no Examples) rather than conceptual. Fix the ship-blocker and the four HIGH findings and this likely scores 68+.