The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, analyze, and refactor shell scripts provided by the user. This creates a surface for indirect prompt injection where malicious instructions embedded in the input code could attempt to override the agent's behavior during the cleanup process.
Ingestion points: The agent reads the entirety of user-provided scripts as specified in commands/cleanup.md under the "Process" section.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the input scripts are provided.
Capability inventory: The skill utilizes the Bash tool to perform syntax checks and execute scripts for baseline verification.
Sanitization: There is no evidence of sanitization or safety filtering applied to the ingested script content before it is processed by the agent.
[COMMAND_EXECUTION]: The cleanup procedure instructs the agent to execute untrusted code to verify behavior.
Evidence: commands/cleanup.md directs the agent to "run the help screen if the script has one" to capture a baseline before refactoring.
Risk: While typically used for verification, executing user-provided scripts (even with help flags) can trigger unintended side effects if the script is maliciously crafted to perform actions beyond displaying documentation.

writing-clean-shell