The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a variety of ecosystem-standard command-line utilities, including npx, pytest, cargo, go, gitleaks, and ast-grep (sg).
[DYNAMIC_EXECUTION]: The skill executes arbitrary project-defined scripts via npm run build and runs project test suites (vitest, jest, pytest, go test, cargo test), which involves executing code found in the local environment. It also executes a short Python script to parse and list scripts from package.json.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from untrusted sources, specifically project configuration files (package.json, pyproject.toml, etc.) and the output logs (stdout/stderr) of the verification tools. Malicious instructions embedded in these files or triggered in error messages could potentially attempt to influence the agent's logic during the verification cycle.

verify