The Agent Skills Directory

[SAFE]: The skill is a set of guidelines and templates for context summarization. It does not contain executable code, external dependencies, or network operations.
[DATA_EXPOSURE]: The instructions guide the agent to include relevant file paths, error messages, and decision logs in its summaries to maintain task continuity. This data is handled within the agent's internal workflow for the purpose of state persistence and is not exfiltrated to external sources.
[PROMPT_INJECTION]: While the skill processes potentially untrusted data such as logs and terminal output (a surface for indirect prompt injection), it uses structured markdown templates for its output, which helps mitigate the risk of the agent misinterpreting summarized content as primary instructions.

context-compression