The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill performs scanning on untrusted repository content and git history, which introduces an attack surface for indirect prompt injection if an attacker embeds malicious instructions within strings or files targeted by the scanner.
Ingestion points: Gitleaks detection output files (gitleaks-report.json, findings.json) and raw git history metadata parsed during scanning.
Boundary markers: The instructions recommend using the --redact flag for log privacy, but they do not provide specific prompt delimiters or instructions to the agent to disregard potential natural language commands found within the scanned data.
Capability inventory: The skill allows the agent to execute shell commands for scanning, perform file operations to install pre-commit hooks, and utilize git commands for history modification.
Sanitization: There are no documented procedures for sanitizing or escaping the content of detected secrets before they are presented to the agent for analysis or remediation planning.
[EXTERNAL_DOWNLOADS]: The skill utilizes official and well-known security tools from trusted sources to perform its primary functions.
It fetches the gitleaks binary via the official Homebrew package manager and legitimate Docker images (zricethezav/gitleaks).
It references the official gitleaks/gitleaks GitHub repository for pre-commit hook configurations and GitHub Actions.
It recommends standard industry tools like git-filter-repo and BFG Repo-Cleaner for secure git history remediation.
[COMMAND_EXECUTION]: The skill relies on standard shell and git command execution to facilitate security auditing and credential lifecycle management.
Primary workflows include repository auditing (gitleaks detect), automated hook installation (.git/hooks/pre-commit), and repository sanitization via git filter-repo. These commands are consistent with the skill's intended purpose as a DevSecOps utility.

secrets-gitleaks