← Back to list
Claude Skill Creator (LLM Council) Checked
Community meta-skill using the LLM Council methodology: interview intent, generate draft, immediate
7.5Overall rating
0Views
0Downloads
1Stars
Installation
🤖 Install via AI
Copy the prompt below and send it to your AI assistant (e.g. Claude Code) — it will follow the instructions and install automatically.
Install the "Claude Skill Creator (LLM Council)" skill by following the instructions at https://skill123.me/install/valorisa-skill-creator.
⌨️ Command line install
Run in your terminal — downloads and installs to ~/.claude/skills/.
curl -fsSL https://skill123.me/install/valorisa-skill-creator.sh | bash
📦 Download ZIP
Download the zip and extract it into your skills directory (e.g. ~/.claude/skills/), then restart your session.
⬇ Download v1.0.0 · 18 KB🛡 Dynamic test
—
Not live-tested yet — this skill is rated on static analysis only.
About this skill
Community meta-skill using the LLM Council methodology: interview intent, generate draft, immediate smoke test, iterate. Emphasizes the minimalist philosophy that a skill is just a markdown file.
Documentation
Community meta-skill using the LLM Council methodology: interview intent, generate draft, immediate smoke test, iterate. Emphasizes the minimalist philosophy that a skill is just a markdown file.
Score breakdown
Trigger
3.0
No frontmatter description at all — the recall mechanism is missing; the body 'Trigger' section lists several phrasings ('create a skill', 'new skill', 'skill for X', 'make me a skill', /skill-creator, build/edit/improve a custom command) but 1.1 scores 0 since there is no functioning description. Moderate phrase variety, limited synonyms, no When-Not boundary. 1.1=0, 1.2=4, 1.3=2, 1.4=0.
Structure
6.7
SKILL.md at 213 lines is fine; templates/ (example outputs) and tests/ (test prompts) are sensibly separated and no redundant README/CHANGELOG. However SKILL.md never mentions templates/ or tests/ — no loading pointers, so progressive disclosure is broken and the files are orphaned. Non-standard directory names (not references//assets/). 2.1=4, 2.2=3, 2.3=1, 2.4=2.
Workflow
9.6
Best-in-batch workflow: 5 phases (interview -> generate -> council validation -> iterate -> finalize+test), Generator+Reviewer pattern with an explicit markdown skeleton, adaptive freedom (skip menu when spec is precise, adapt to non-Claude LLMs), strong guards (council fallback checklist, Windows symlink warning, skip-if-clear), and a genuine verification loop (council review, 7-point checklist, fresh-session test). 3.1=6, 3.2=6, 3.3=5, 3.4=4, 3.5=3.
Content
9.0
Concise imperative instructions, no filler; generation template is concrete. Only one runnable command (the optional git clone/cp install snippet with full args); the rest is textual instruction, appropriate for a generator. 4.1=4, 4.2=2, 4.3=3.
Engineering
6.0
Fatal engineering flaw: SKILL.md lacks YAML frontmatter entirely — no name, no description — so the skill is not machine-indexable and cannot auto-trigger (5.1=0). No scripts (N/A). No dangling references (llm-council is explicitly declared not bundled), though orphaned templates//tests/ reflect incomplete wiring. 5.1=0, 5.2=4, 5.3=2.
Security
10.0
Capability manifest: data reads = existing skill files the user asks to improve; side effects = creates ~/.claude/skills/[name]/ and writes SKILL.md there; network = optional, explicit git clone from github.com/tenfoldmarc/llm-council-skill (trusted domain, fully disclosed, not bundled to avoid version drift); credentials = none; persistence = none. Capabilities match declared scope. The generated-skill template embeds a safety block requiring confirmation before destructive ops — positive safety posture. No injection, no secrets, no scripts. No veto.
A well-designed generator skill (213-line SKILL.md) that interviews the user with a 7-question menu, generates a skill from a markdown skeleton, validates it via an 'llm-council' peer review (with a structural checklist fallback), iterates, and finalizes with a fresh-session test. Bundles 5 example templates and 5 test prompts. Excellent workflow design: adaptive interview (skips menu when spec is complete), explicit boundaries/autonomy questions, safety notice baked into the generated template, and a real verification loop. Two significant defects: (1) SKILL.md has NO YAML frontmatter — no name or description — so as shipped it cannot be indexed or auto-triggered by Claude Code, gutting recall; (2) the bundled templates/ and tests/ directories are never referenced from SKILL.md, leaving them orphaned. Security is clean: capabilities (write to ~/.claude/skills/, optional git clone of a GitHub skill repo) match declared scope, no injection, no secrets, no credential access, no persistence. No veto.
Assessment methodology
AI-assessed
Evaluated by automated analysis across quality, usability, and security dimensions. This is not a substitute for your own review — always inspect a skill before installing it.
Version history
v1.0.0
2026-10-13