The Agent Skills Directory

[PROMPT_INJECTION]: No malicious instruction overrides or safety bypass attempts were found. The skill uses imperative language to enforce a design workflow, which is standard for process-guiding skills.
[DATA_EXFILTRATION]: No network exfiltration or sensitive file access was identified. File operations are confined to reading project documentation and writing design artifacts to paths like docs/designpowers/briefs/ and design-state.md.
[COMMAND_EXECUTION]: No shell commands, subprocess calls, or script execution patterns are present in the skill instructions.
[REMOTE_CODE_EXECUTION]: The skill does not perform any external downloads or execute remote scripts.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input and existing project files (Ingestion) to generate markdown design artifacts (Capability). While it lacks explicit boundary markers or sanitization, the risk is minimal as the output is descriptive documentation and not executable code.

design-discovery