cartooli

mega-eval

A Claude skill that runs a comprehensive, multi-phase evaluation of any idea, product, or feature set. Feed it text, docs, URLs, or any combination — it produces 6 structured Word documents covering critical feedback, competitive landscape, strengths, design issues, action items, and content strategy.

What it does

INPUT (text / doc / URL / combo)
     │
     ▼
Phase 0: Ingest & normalize all inputs into an Evaluation Brief
     │
     ├── Phase 1A: Hater Mode (12-persona critical feedback)
     ├── Phase 1B: Competitive & Market Landscape (web research)
     └── Phase 1C: Strengths & Opportunities
            │
            ▼
Phase 2: Synthesis — critical fixes, design issues, next steps
     │
     ▼
Phase 3: Content Strategy Outline
     │
     ▼
Phase 4: Assemble 6 deliverable .docx files

Deliverables

File	Contents
00-executive-summary.docx	Standalone decision-maker document with verdict, top findings, and prioritized actions
01-hater-mode-feedback.docx	Simulated critical feedback from 12 internet audience personas
02-competitive-landscape.docx	Competitors, market size, positioning analysis, risks
03-strengths-opportunities.docx	What's working, growth angles, unfair advantages, ideal customer profiles
04-critical-fixes-and-design.docx	Prioritized fixes, design inconsistencies, non-breaking next steps
05-content-strategy-outline.docx	Long-form outline for publicly positioning/pitching the idea

Where it runs, privacy, and local files

• Host environment: Mega-eval targets Claude Code (skills + hater-mode, long-form-outline, docx, and parallel subagents). Other Claude apps or agents may run the steps partially or need you to follow SKILL.md by hand—do not assume identical behavior.
• Sensitive data: You may pass URLs, pitch text, or confidential documents into the pipeline. Treat generated .docx files, raw markdown intermediates, and optional run-log.md as sensitive unless you redact them. To skip writing a run log, set MEGA_EVAL_LOG=off (see Run feedback).
• PDF / Office inputs: Extracting text from local .pdf, .docx, or .pptx via scripts/ingest.py requires pdftotext and pandoc (optional tools below). Plain text, Markdown, pasted content, and many URLs work without them.

Install

Claude Code plugin install (recommended)

Requires: Claude Code 1.0.33 or later. Run claude --version to check.

Security note: Only add marketplaces from sources you trust — plugins run with your user privileges.

Install the full pipeline and all phase skills via the Claude Code plugin system:

# Step 1 — register the marketplace (use owner/repo, NOT a raw URL)
/plugin marketplace add Cartooli/mega-eval-skill

# Step 2 — install the plugin
/plugin install mega-eval@cartooli

# Step 3 — activate without restarting
/reload-plugins

That installs the full pipeline skill (mega-eval) plus all seven phase-only skills (mega-eval-brief, mega-eval-competitive, etc.) in one shot.

Why Git-based add (owner/repo) and not a direct URL? The plugin tree uses relative paths to a shared references/ directory. Relative paths only resolve when the marketplace is added via Git — a direct URL to the raw marketplace.json file cannot resolve them. Always use the owner/repo or git URL form.

Plugin not appearing after install? Run /reload-plugins first. If still missing, clear the cache and reinstall:

rm -rf ~/.claude/plugins/cache
# then re-run /plugin marketplace add ... and /plugin install ...

Manage the marketplace:

/plugin marketplace update cartooli   # pull latest version
/plugin marketplace remove cartooli   # remove marketplace + uninstalls its plugins
/plugin uninstall mega-eval@cartooli  # uninstall plugin only

Third-party marketplaces (like this one) have auto-update disabled by default. Run /plugin marketplace update cartooli after new releases, or enable auto-update in the plugin UI (/plugin → Marketplaces → cartooli → Enable auto-update).

---

Full pipeline only (mega-eval)

Copy the skill folder into your project:

# Clone this repo
git clone https://github.com/Cartooli/mega-eval-skill.git

# Copy into your Claude Code skills directory (include the whole references/ folder for learnings + prompts)
cp -r mega-eval-skill/SKILL.md mega-eval-skill/references mega-eval-skill/scripts ~/.claude/skills/mega-eval/

Or if you prefer, copy just the skill folder contents into any .claude/skills/mega-eval/ directory in your project.

Phase-only skills (thin entrypoints)