The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/zoom_out_inventory.py uses the subprocess module to execute git commands, specifically git rev-parse and git ls-files, to locate the repository root and list files for scanning. While it uses a list for arguments to mitigate shell injection, the path is derived from user-supplied input.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes content from the repository being inspected, which may contain malicious instructions designed to manipulate the agent's behavior.
Ingestion points: The safe_read function in scripts/zoom_out_inventory.py reads various files, including source code and configuration files like .env, to find matches for the target terms.
Boundary markers: The instructions do not specify any delimiters or safety prompts for the agent to distinguish between its instructions and the data it reads from the repository.
Capability inventory: The agent has the capability to execute shell commands, read/write files, and perform further analysis based on the inventory results.
Sanitization: The script implements basic sanitization by reading only the first 250KB of files and truncating individual line matches to 180 characters in the output.

zoom-out