Secure an app
with agent skills.

Start with the part of the application you need to examine: authentication, input handling, dependencies or deployment exposure. These skills provide workflows and checks; installing one does not establish that an application is secure.

Filter these skills by stack and access

Choose a skill

Match the skill to your application and the review you have authorized. A pre-launch checklist, a framework-specific review and a threat model have different scopes. Compare what each skill can read, change or send over the network.

Explore the launch week kit

Before you install

Define the repository, local environment and permitted test targets. Use test accounts and synthetic data. Keep credentials out of prompts and reports, and ask before testing third-party systems or changing production settings.

Open a listing to read its source, access labels and review evidence before copying its install command. The automated scan covers SKILL.md, not referenced files or runtime behavior. Read the review limits.

Skills for this task

19 available skills, listed A to Z. Task labels help narrow the choice; they do not establish fit for your project.

Check the result

  1. 01

    Trace the affected input and permission boundaries in the code rather than relying only on scanner output.

  2. 02

    Use safe negative tests for the specific risk, such as rejected input or access by the wrong test account.

  3. 03

    Verify each approved fix, record remaining limitations and keep deployment approval separate from a successful local check.