The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled measurement script (scripts/measure.py) using uv run python. This script is used locally to calculate token counts and identify optimization opportunities without performing any network operations.
[DATA_EXFILTRATION]: The skill accesses local configuration files within ~/.claude, as well as project-specific CLAUDE.md and MEMORY.md files. This access is necessary for the primary purpose of token optimization. No data is transmitted externally; results are displayed only to the user.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes user-controlled local files (Category 8).
Ingestion points: The scripts/measure.py script reads the content of agent descriptions and local project files.
Boundary markers: Absent; the content is processed as raw text for tokenization and translation suggested by the agent.
Capability inventory: The agent is instructed to modify local configuration files based on the measurement report.
Sanitization: Absent; however, the skill mandates a protocol where the user must approve all changes, and the agent is instructed to preserve semantic meaning.

token-optimize