The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external repositories, specifically README files and source code, which can be manipulated to influence agent behavior. \n- Ingestion points: Repository README files and source code are read during the audit process (documented in SKILL.md). \n- Boundary markers: No explicit markers or instructions are provided to distinguish between content to be audited and commands to be ignored if they attempt to manipulate the agent. \n- Capability inventory: The skill leverages the agent's ability to execute shell commands and scripts found within the README files to verify installation and usage instructions (SKILL.md, Step 4). \n- Sanitization: No sanitization or validation of the commands extracted from README files is mentioned before they are executed. \n- [DYNAMIC_EXECUTION]: The skill's verification process involves executing arbitrary shell commands and code snippets parsed from external, untrusted sources. \n- Evidence: SKILL.md Step 4 instructs the agent to "Execute the documented install command exactly as written" and "the first usage example exactly as written." This behavior is central to the skill's purpose but allows for the execution of potentially malicious code embedded in a repository's documentation.

readme-optimization