The Agent Skills Directory

[COMMAND_EXECUTION]: The skill incorporates procedural instructions to use the local git binary to retrieve repository metadata for its documentation tasks.
Evidence in guides/02-direct-invocation.md specifies that the agent should shell out to git log and git blame if the primary TypeScript driver is unavailable.
These operations are limited to read-only metadata retrieval and are consistent with the skill's primary purpose.
[PROMPT_INJECTION]: As a documentation tool, the skill has an inherent surface for indirect prompt injection by processing external source code and commit messages.
Ingestion points: Source code from repository chunks and commit messages retrieved from git logs (identified in guides/01-canonical-invocation.md and guides/07-adr-detection.md).
Boundary markers: The skill utilizes a set of predefined markdown templates (found in the templates/ directory) that encapsulate extracted code and signatures within markdown code blocks.
Capability inventory: The skill is capable of writing markdown files to the library/knowledge-base/wiki/ directory and executing local git commands.
Sanitization: The extraction process uses specialized AST parsers (such as ts-morph and react-docgen-typescript) to isolate specific code units, which provides structural isolation for the processed data.

wiki-weapon