The Agent Skills Directory

[SAFE]: The skill is primarily instructional, providing guidelines for agents to handle large data efficiently and reduce token costs.
[SAFE]: All helper scripts are implemented using the Python 3 standard library with no external dependencies or network access detected.
[SAFE]: The local caching mechanism (CCR) uses SHA256 hashes for filenames and includes regex validation to prevent directory traversal when the agent retrieves original files.
[SAFE]: The skill includes explicit security and privacy guidance in its documentation, advising users to exclude the cache directory from version control and explaining the risks of local data storage.
[SAFE]: No prompt injection or behavior override patterns were detected; instructions focus on task efficiency while maintaining safety and accuracy invariants, such as never dropping error information.
[SAFE]: The skill creates a potential attack surface for indirect prompt injection by processing external data (logs, search results), but it includes boundary markers and explicit rules for the agent to distinguish between tool results and human instructions.

headroom