The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading local source code files to identify mutation targets, creating a surface where malicious code comments or strings could attempt to influence the agent's behavior. 1. Ingestion points: The read_file tool is used in SKILL.md (Step 3) to scan target source files. 2. Boundary markers: The instructions do not specify any delimiters or 'ignore' warnings for the content of the files being read. 3. Capability inventory: The skill has access to run_in_terminal, edit_file, and create_file. 4. Sanitization: There is no mentioned sanitization or filtering of file content during the identification of semantic constructs.
[COMMAND_EXECUTION]: The skill uses run_in_terminal to execute test suites based on local project configuration (e.g., npm test, go test, pytest). While necessary for the skill's purpose, this grants the agent the ability to execute commands defined within the user's local project files.

mutation-hunter