The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: User-provided input from 'user_request' and 'context' in 'src/index.js' is used to construct the core purpose and contents of new SKILL.md files. Boundary markers: The skill does not employ delimiters or 'ignore' instructions when interpolating untrusted data into the generated drafts. Capability inventory: The skill environment includes validation scripts in 'scripts/' capable of subprocess execution. Sanitization: No sanitization is performed on the primary 'purpose' string before it is embedded into the output artifacts.
[COMMAND_EXECUTION]: The skill performs dynamic command execution. Evidence: The script 'scripts/validate_reasoning_artifacts.py' uses 'subprocess.run' to execute a Python script located at a computed relative path ('../../shared/scripts/validate_artifact_types.py'). This facilitates the validation of reasoning artifacts using shared vendor utilities.

ai-architect