The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (source code and git history) which could potentially contain malicious instructions intended to mislead the agent.
Ingestion points: Reads source code, configuration files, and git commit history through the /secrets-scan command.
Boundary markers: None explicitly defined, but the output is restricted to a specific, redacted format.
Capability inventory: Includes file system reading, regex-based analysis, and the ability to suggest or execute shell commands for remediation.
Sanitization: Implements a mandatory redaction rule where credential values are masked (e.g., AKIA****MPLE) or replaced with (redacted) before being displayed to the user.
[COMMAND_EXECUTION]: The skill documentation includes shell commands for secret remediation and prevention.
Evidence: References to git filter-branch, bfg, and npx husky are provided as remediation steps in SKILL.md.
Context: These are industry-standard tools for the stated purpose. The skill includes a mandatory approval gate, stating that no remediation actions may be executed without explicit user consent for that specific action.

secrets-scan