The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by ingesting content from untrusted source code and interpolating it into CLAUDE.md as project-level rules.
Ingestion points: All source files identified during the scan phase (SKILL.md Step 2.4), specifically documentation comments and architectural patterns.
Boundary markers: No explicit sanitization or delimiters are mentioned to isolate extracted code comments from the generated instructions in CLAUDE.md (SKILL.md Step 2.6).
Capability inventory: The skill has the capability to modify the agent's root instruction file (CLAUDE.md), perform git operations (git push, git clone), and create/modify local files.
Sanitization: The skill lacks validation or filtering for content extracted from source code before placing it into the project's rule definitions.
[COMMAND_EXECUTION]: The skill uses shell commands for core functionality, including git operations for diffing (git diff), state management (git rev-parse), and branching (git checkout).
[EXTERNAL_DOWNLOADS]: The Docusaurus sync feature (SKILL.md Step 5) involves cloning and pulling from remote git repositories defined in the user configuration. While the destination is user-controlled, this involves executing remote operations based on configuration data.

ws-codebase-documenter