The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a suite of Python scripts (e.g., smart_crusher.py, log_compressor.py) located in the scripts/ directory. These scripts are used for content selection and compression. They are authored by the skill provider and use only standard Python libraries with no external dependencies.
[SAFE]: The skill implements a local data persistence mechanism called CCR (Compress-Cache-Retrieve). Before compression, original content is stored in a hidden directory (.headroom-cache/) using SHA-256 hashes as keys. The analysis confirms this cache is stored locally and is not exfiltrated to any remote server.
[SAFE]: The skill uses clear instructions and boundary markers to manage agent behavior. It provides explicit rules for preserving errors and maintaining the integrity of system prompts and user turns. There are no attempts to override safety guidelines or bypass security filters.

headroom