The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/scan-codebase.py) to perform local file system traversal and token counting. This is a standard operational step for the skill's stated purpose.
[EXTERNAL_DOWNLOADS]: The skill identifies tiktoken as a dependency. This is a standard and trusted library used for token estimation. The instructions provide standard installation commands for this package.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes the content of arbitrary codebase files.
Ingestion points: Local files are read by the scanner script and the subagents during analysis in SKILL.md and scripts/scan-codebase.py.
Boundary markers: The subagent instructions in SKILL.md (Step 4) do not define specific delimiters for separating project code from instructions.
Capability inventory: The skill can read local files, write documentation to the filesystem, and spawn subagents via the Task tool.
Sanitization: No sanitization of the files being documented is performed.

cartographer