The Agent Skills Directory

[SAFE]: The skill is primarily a documentation and design framework. It does not perform network operations, file system modifications (writes), or code execution. Its primary purpose is to help an agent process information and generate structured UX reports based on local repository context.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates high security awareness regarding untrusted content. In references/agentic-setup.md, it establishes an 'Untrusted content boundary' that explicitly instructs the agent to treat repository, tracker, and document content as data only. It specifically commands the agent to refuse and report common prompt injection patterns like 'ignore previous instructions' or 'run this command' if encountered in project files. This mitigation addresses the attack surface of reading external data (Category 8).
[CREDENTIALS_UNSAFE]: The skill includes a 'Secrets hygiene' rule in references/rules.md, which prohibits the inclusion of secrets, tokens, or credentials in any reports or specifications generated by the tool.
[PROMPT_INJECTION]: A static detector hit for PI_IGNORE_INSTRUCTIONS was identified as a false positive. The phrase 'ignore previous instructions' appears in references/agentic-setup.md not as a directive to the agent, but as an example of a malicious pattern that the agent is instructed to detect, refuse, and report.

om-ux-shape