The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands to verify the local development environment and check project state.
Evidence: SKILL.md uses git rev-parse, node, pnpm, pwd, and ls to identify the project root and confirm the availability of required scanning tools.
[INDIRECT_PROMPT_INJECTION]: The skill establishes the entry point for an analysis pipeline that ingests untrusted codebase data into the agent context.
Ingestion points: The skill targets the current working directory (pwd) to begin reverse-engineering the codebase.
Boundary markers: No delimiters or explicit 'ignore embedded instructions' markers are defined for the content being scanned.
Capability inventory: The skill has access to Read, Glob, and Bash tools, including a source capability which allows for the execution of script content, creating a significant attack surface if malicious code is encountered.
Sanitization: No sanitization or validation logic is present at this orchestration level to filter potentially malicious instructions within the project files.

uw-start