The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the agent to produce and execute code at runtime to validate the behavior of the application code being analyzed. Evidence: 'Write a script or test that runs the real code, run it, and paste what happened' (SKILL.md, Step 5).
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad surface for indirect injection because it reads untrusted externally-provided code (diffs and PRs) and has the permission to execute scripts based on its interpretation of that code. Ingestion points: It reads diffs, PRs, and commit history from the development environment (SKILL.md, Step 1). Boundary markers: The instructions do not mandate the use of delimiters or warnings to ignore embedded instructions within the analyzed diffs. Capability inventory: The agent can write and execute arbitrary scripts or tests that import local libraries (SKILL.md, Step 5). Sanitization: There are no instructions to sanitize or validate the content of the diffs before they are processed for verification code generation.
[COMMAND_EXECUTION]: The agent is expected to execute local commands and scripts to perform symbol searches and runtime verification of the codebase (SKILL.md, Step 3 and 5).

blast-radius