The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines hooks (PreToolUse, PostToolUse, and Stop) that execute shell commands to maintain an audit log in the system's temporary directory. It also provides a CLI entry point via python -m conserve.context_analyzer to analyze current context usage.
[INDIRECT_PROMPT_INJECTION]: The skill's assessment modules ingest conversation history, user messages, and task descriptions to calculate token usage and reasoning complexity. This analysis of untrusted data is the primary mechanism for its context optimization and delegation logic.
[EXTERNAL_DOWNLOADS]: The skill documentation and theory modules reference external Python packages conserve and leyline for context monitoring and pressure calculations. These are treated as local dependencies for the skill's analytical operations.

context-optimization