The Agent Skills Directory

[COMMAND_EXECUTION]: Executes standard development CLI tools (npm, pnpm, pyright, ruff, git) to perform project validation. These operations are within the expected scope of a verification skill and are restricted to the local development environment.
[DATA_EXPOSURE]: Includes a security scan phase that proactively searches the local codebase for potential secret leaks (e.g., sk- or api_key patterns) to warn the user. This is a defensive security feature aimed at preventing accidental exposure of credentials.
[SAFE]: The skill does not perform network operations, use any form of obfuscation, or attempt to establish persistence. All actions are transparent and aligned with the stated purpose of code quality assurance and verification.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from build tools and test suites which could technically contain untrusted data if an attacker controls the source code being verified.
Ingestion points: Command outputs from npm, pnpm, pyright, ruff, and git (SKILL.md).
Boundary markers: No explicit delimiters are used in the prompt instructions.
Capability inventory: Shell command execution via project-standard build and lint tools.
Sanitization: Not applicable as the skill primarily reports on command exit codes and log snippets.

verification-loop