← Set up FeedLab with AI

Skills

A skill is a short instruction file that teaches your AI agent how to do one job well with FeedLab: what to look for, how to check its work, and when to ask you before writing anything. They work alongside the FeedLab MCP server, which you'll need connected first.

Design tests

feedlab-design-tests

Turns your code into clear manual test cases, organised into suites and groups.

  • Maps every screen, action, rule, and role in the area you name
  • Writes cases a tester can follow: preconditions, steps, expected result, priority
  • Updates or archives existing cases when the code has changed
Ask
Design tests for user authentication and add them to my project in FeedLab

Result: New test cases under Tests, ready to run in a session.

Read the skill file →

Review code

feedlab-review-code

Finds real deficiencies in your code, verifies them, and registers them as issues.

  • Looks for bugs, reliability, performance, accessibility, and maintainability problems
  • Traces each one to confirm it before reporting, with file and line
  • Rates severity and suggests a fix, then marks issues resolved once fixed
Ask
Review this codebase for bugs and reliability problems and register them as issues in my project in FeedLab

Result: Issues in the project's feedback, tagged code-review.

Read the skill file →

Security review

feedlab-security-review

Audits your code for vulnerabilities and registers the confirmed ones as security issues.

  • Maps your attack surface: entry points, auth checks, sensitive data
  • Checks access control, injection, auth and sessions, secrets, data exposure, and more
  • Proves each finding with a concrete attack, and never puts secrets in an issue
Ask
Run a security review of this codebase and register the vulnerabilities as issues in my project in FeedLab

Result: Security issues in the project's feedback, tagged code-review and security.

Read the skill file →

Install for Claude Code

Run this once in your terminal. It saves every skill to ~/.claude/skills, so they're available in all your projects. Claude Code picks the right one from what you ask; restart it after installing.

mkdir -p ~/.claude/skills/feedlab-design-tests && curl -fsSL https://feedlab.cloud/skills/feedlab-design-tests/SKILL.md -o ~/.claude/skills/feedlab-design-tests/SKILL.md
mkdir -p ~/.claude/skills/feedlab-review-code && curl -fsSL https://feedlab.cloud/skills/feedlab-review-code/SKILL.md -o ~/.claude/skills/feedlab-review-code/SKILL.md
mkdir -p ~/.claude/skills/feedlab-security-review && curl -fsSL https://feedlab.cloud/skills/feedlab-security-review/SKILL.md -o ~/.claude/skills/feedlab-security-review/SKILL.md

To install for one project only, change ~/.claude/skills to .claude/skills and run it in the project folder. Run the command again any time to get the latest versions.

Other agents

Skills are plain Markdown, so any agent that accepts custom instructions can use them. Open a skill file above and add its contents to your agent's rules or instructions, for example a Cursor rule or a system prompt, alongside the FeedLab MCP server.

Good to know

  • Every skill shows you its plan or findings and waits for your approval before adding anything to FeedLab.
  • Reviews file issues in batches, and your team gets one notification per review.
  • The security review describes each attack in words and never copies secrets or personal data into an issue.
  • Agents make mistakes. Read what they propose the way you'd review a teammate's work.