The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the codebase to derive test expectations.\n
Ingestion points: The instructions specify reading docstrings, README files, issue trackers, and existing code usage (SKILL.md).\n
Boundary markers: No specific boundary markers or sanitization logic is defined for these inputs to distinguish instructions from data, which could allow malicious content in a README or docstring to influence agent behavior.\n
Capability inventory: The skill has access to Bash, Write, Edit, and Read tools, allowing it to modify and execute code based on these inputs.\n
Sanitization: No sanitization or escaping of external content is mentioned before it is interpolated into test generation logic.\n- [DYNAMIC_EXECUTION]: The skill generates new test scripts and subsequently executes them.\n
The cover mode involves writing new test files (SKILL.md section 'Write the tests') based on the agent's interpretation of the codebase.\n
These scripts are executed via the repository's test runner, which is invoked through the Bash tool.\n- [COMMAND_EXECUTION]: The skill identifies and executes shell commands defined within the project's configuration files.\n
The done-gate and runner setup phases involve discovering and running commands from package.json, Makefile, pyproject.toml, and other build system files, which could lead to execution of malicious local scripts if the analyzed repository is compromised.

testkit