The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requests Bash permissions to run security utilities for identifying exposed secrets and verifying configurations. It specifically references standard tools like git secrets and trufflehog, which is consistent with its stated purpose of security auditing.- [DATA_EXFILTRATION]: While the skill contains procedures to find sensitive files (e.g., SSH keys, .env files), it does not include any commands or network requests that would transmit this data externally.- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted codebase data during audits. Ingestion points: the skill reads files using cat and find in references/audit-tools.md. Boundary markers: the instructions do not include specific delimiters or warnings to ignore instructions within audited files. Capability inventory: the skill is granted Bash, Write, and Edit tools in SKILL.md. Sanitization: no specific input validation is defined. This surface is considered a necessary part of the auditing functionality.- [EXTERNAL_DOWNLOADS]: The skill contains references to official compliance and security documentation from cursor.com. These are legitimate links to well-known service resources and do not involve unverified downloads.

cursor-compliance-audit