The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the ripgrep (rg) binary through subprocess.run to facilitate text search across the codebase. It uses a list for arguments and avoids shell execution, mitigating standard command injection risks.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It ingests content from files within the project directory and presents them to the agent. If a codebase contains malicious instructions, the agent could potentially follow them when interpreting search results.
Ingestion points: scripts/search/nodes/engines.py and scripts/smart_ast/engine.py read file contents from the project directory.
Boundary markers: Search results are encapsulated in XML tags with <![CDATA[ ... ]]> blocks in scripts/search/nodes/formatter.py.
Capability inventory: The skill has read access to the local filesystem and the ability to execute the ripgrep subprocess.
Sanitization: Output is XML-escaped, but the content of the matched code snippets is not sanitized for malicious natural language instructions.

code_tools