The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to estimate file counts and verify applied fixes. Step 0a in sweep-guide.md uses git ls-files and find. Step 2e executes project-specific commands such as npm test, pytest, cargo test, or go test. While this is expected behavior for a developer-oriented tool, executing commands derived from project configuration files (like package.json) presents a risk if the codebase being scanned is untrusted.
[INDIRECT_PROMPT_INJECTION]: The skill has a high exposure to indirect prompt injection because it reads the entire project scope while having the capability to modify files and execute commands. Malicious instructions hidden in code comments or strings could potentially manipulate the agent's behavior during the 'hands-free' autonomous execution.
Ingestion points: Full repository scanning is performed across four analysis dimensions in sweep-guide.md (Steps 2–5).
Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from obeying instructions found within the code being analyzed.
Capability inventory: The skill uses file modification tools (Edit, Write) and shell tools to perform its primary function.
Sanitization: Content from files is processed without explicit sanitization or filtering to remove potential injection patterns.

brooks-sweep