The Agent Skills Directory

[SAFE]: No malicious code components, remote script downloads, or unauthorized network operations were detected in the skill definition.
[DATA_EXPOSURE]: The skill is instructed to analyze project files, including sensitive areas like runtime configuration and CI/CD pipelines. However, the instructions focus on extracting high-level architectural decisions and 'ownership' layers rather than raw credentials or secret values.
[PROMPT_INJECTION]: The skill includes directives to conceal its AI identity (e.g., 'NEVER mention AI') and adopt a specific persona ('We chose...') to facilitate professional documentation. This is a functional persona constraint rather than a malicious attempt to bypass safety filters or ignore prior instructions.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted codebase files and lacks explicit boundary markers or sanitization logic to separate codebase content from its own instructions. This creates a theoretical surface where malicious comments in a codebase could influence the output. Ingestion points: files within the target directory specified in $ARGUMENTS or the working directory. Boundary markers: absent. Capability inventory: read-only codebase analysis and text generation. Sanitization: absent.

spectra-own