The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the Gitleaks binary from its official GitHub repository (github.com/gitleaks/gitleaks). This is a well-known service. The script verifies the integrity of the downloaded archive using pinned SHA-256 checksums for Linux and macOS platforms before installation.\n- [COMMAND_EXECUTION]: The script executes git commands to analyze repository history and manage git hooks. It also uses curl for the initial binary download and tar for extraction. These actions are limited to the skill's functional requirements for local repository scanning.\n- [DYNAMIC_EXECUTION]: The bash script uses an embedded Python block to process and sanitize JSON reports. This local execution is used to redact secrets and ensures that only non-sensitive metadata is preserved in the final report.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from local git history and diffs.\n
Ingestion points: Reads repository commits and staged changes via gitleaks and git commands (scripts/secret-scan.sh).\n
Boundary markers: Findings are strictly delimited within a JSON structure.\n
Capability inventory: Subprocess execution for gitleaks and git, local file writes for reports and hooks (scripts/secret-scan.sh).\n
Sanitization: The skill explicitly redacts the secret content, matches, and source lines, keeping only metadata fields (RuleID, File, Line, Commit, Fingerprint).\n- [SAFE]: The skill demonstrates several security best practices, such as failing closed if the scanner is missing or version-mismatched, enforcing that allowlists/ignores are committed for auditability, and setting report file permissions to 0600 to prevent unauthorized access.

private-secret-scanning