The Agent Skills Directory

[SAFE]: The skill implements strong privacy controls through a dedicated anonymization protocol. It explicitly forbids sending raw financial data, customer names, or employee names externally, requiring the agent to use relative ranges or generic descriptors instead.
[SAFE]: Access to the local file ~/.claude/company-context.md is limited to the skill's primary function of managing session context. The skill incorporates user confirmation steps before modifying this file, preventing unauthorized data persistence or modification.
[SAFE]: The staleness checks based on the Last updated field are used purely for data quality management and do not gate any suspicious network or system operations.

context-engine