agentsclimarketplace

Gemini skill creator

Skill grantmcd/skills/gemini-skill-creator

High-quality, objectively evaluated, and secure AI agent skills.

Install
npx -y skills add grantmcd/skills --skill gemini-skill-creator

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use this meta-skill to design, architect, and objectively evaluate high-quality specialized skills following the Agent Skills open standard. Activate whenever the user wants to 'create a skill', 'automate a workflow', or 'refine an agent capability', even if they don't explicitly mention 'Agent Skills' or 'SKILL.md'. Mandates sandboxed evaluation, objective metrics, and interactive review dashboards.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.2 KB, as published. Nobody here has run it

gemini-skill-creator

A meta-skill for designing, testing, and objectively evaluating high-quality specialized skills for the Gemini CLI and other AI agents.

Core Mandate

This skill's purpose is to move beyond "vibe-based" skill creation. It enforces the Agent Skills open standard and follows a rigorous cycle of: 0. Specification Refresh: ALWAYS start by retrieving the latest specification from https://agentskills.io/specification using web_fetch to ensure total compliance with the latest standard.

  1. Intent Capture: Understanding the precise goal and target audience.
  2. Pattern Discovery: Researching existing codebase conventions.
  3. Sandboxed Evaluation: ALL drafting and testing MUST occur in an isolated temporary directory (e.g., .gemini/tmp/evals/<skill-name>) to prevent workspace pollution.
  4. Interactive Review: Providing a rich, visual dashboard for human-in-the-loop validation before changes are committed.

IMPORTANT: Always review the latest specification and best practices at agentskills.io before finalizing a skill.

1. Intent & Research

When a user wants to build a skill, don't just draft it. First, define:

  • Success Criteria: What specifically makes an output "correct"?
  • Triggers: When exactly should this skill be activated?
  • Pattern Validation: Use codebase_investigator to find local architectural styles before drafting.

2. Drafting the SKILL.md

Follow these principles and the official Specification:

  • Directory Structure: Every skill MUST be its own directory containing a SKILL.md file. Use subdirectories like scripts/, ui/, and references/ for supporting artifacts.
  • File Size: Keep SKILL.md under 500 lines and ~5,000 tokens. Move heavy reference material to the references/ folder to ensure context efficiency.
  • Imperative Instructions: Use clear, direct commands.
  • The "Why": Explain the reasoning behind instructions to improve agent compliance.
  • Security-First: MUST include a mandate for protecting credentials and sensitive data.
  • Tool-Specific Guidance: Explicitly mention how to use Gemini-native tools like grep_search, generalist, and mcp_context7.

3. Automated Evaluation Framework

Every skill must be verified before deployment.

  1. Isolation: Create a clean, temporary workspace (e.g., in .gemini/tmp/evals/) for the evaluation run.
  2. Define Evals: Create evals/evals.json with 3-5 diverse test prompts and specific assertions.
  3. Parallel Execution: MUST use the generalist sub-agent to run these prompts in parallel within the sandbox and capture performance metrics.
  4. Capture Timing/Tokens: Record the execution time and token usage for each eval to identify performance bottlenecks.

4. The Review Dashboard

To ensure high-quality delivery:

  • Render UI: Run python3 scripts/generate_review.py results.json ui/viewer.html review.html to generate an interactive dashboard.
  • Human Approval: Share the path to review.html. The user MUST review the outputs and assertions before the skill is considered "production ready."

5. Publishing & Discovery

  • Description Tuning: Optimize the description in frontmatter for accurate triggering.
  • skills.sh Integration: Follow the root-level directory structure (my-skill/SKILL.md) for compatibility with npx skills add.

Reporting & Feedback

ALWAYS provide a summary of the evaluation results:

  • Pass Rate: % of assertions that passed.
  • Efficiency: Token usage and timing metrics.
  • Review Path: Link to the generated review.html for human inspection.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.