The Agent Skills Directory

[SAFE]: The skill is a template for performing security analysis and does not execute arbitrary code or exfiltrate data. The libexec scripts referenced (e.g., raptor-normalize-context-map) are documented as part of the execution lifecycle for formatting results.
[DATA_EXPOSURE]: The skill is designed to read local codebases for analysis, which is its primary stated purpose for security research. It produces JSON reports (context-map.json, variants.json) intended for a validation pipeline. No evidence of sending this data to external or untrusted domains was found.
[COMMAND_EXECUTION]: The skill uses a !command syntax in documentation contexts and references internal tools like raptor-sandbox-observe for optional runtime analysis. These are part of the intended developer/researcher workflow and are not used for malicious persistence or privilege escalation.

code-understanding