The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted files (code, config, environment files) from a target directory, creating a potential surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: The skill uses Glob and Grep to locate and read files like .env, .pem, and various configuration files as defined in 'Step 1: Discovery'.
Boundary markers: The skill includes a 'Prompt Injection Safety Notice' section that explicitly instructs the agent to treat all file content as untrusted data and ignore any embedded instructions or commands.
Capability inventory: The skill is restricted to Read, Grep, and Glob tools to inspect repository content, minimizing the impact of any potential injection.
Sanitization: No programmatic sanitization is defined; the skill relies on natural language instructions to the LLM to ensure safety and prevent the extraction or logging of actual secret values.

secrets-management