The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates artifacts based on user-provided feature descriptions and intent. Ingestion points: User-supplied feature descriptions via slash commands. Boundary markers: Absent. Capability inventory: Writes to specs/ and .ai/ directories and executes local analysis tools. Sanitization: Absent. Although it processes untrusted input, the functionality is limited to workspace document generation and standard coding tasks with no evidence of malicious redirection.
[COMMAND_EXECUTION]: The skill instructs the agent to run local analysis commands such as tailor analyze and a specific workspace script at skills/reuse-first/scripts/find-duplicates.js. These are legitimate development tools used for auditing existing code for reuse and do not trigger remote downloads or sensitive data exfiltration.

spec-driven-dev