The Agent Skills Directory

[COMMAND_EXECUTION]: The analyze.js script executes multiple shell commands to inspect the repository state, including git diff, git status, and git branch. It also spawns child processes to run a local review-state checker script (scripts/review-state.js).
[DYNAMIC_EXECUTION]: The analyze.js script uses a dynamic path resolution mechanism for _pluginRoot to load internal modules (utils, feature-resolver, request-status) via require. This resolution depends on environment variables (PLUGIN_ROOT) or runtime directory traversal, which is a form of dynamic loading from computed paths.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository, including filenames and the content of markdown files in the docs/features/ directory.
Ingestion points: analyze.js reads file paths from git status output and reads file content from docs/features/**/*.md to parse status and acceptance criteria.
Boundary markers: No explicit delimiters or boundary markers are used when processing the content of these files.
Capability inventory: The skill can execute subprocesses via analyze.js and can invoke further agent commands through the Skill tool as defined in the Dispatch Mode section of SKILL.md.
Sanitization: The skill attempts to sanitize arguments for its auto-execution mode by filtering tokens to only include those that start with --, include a /, or include a . (intended to allow only flags and paths). However, malicious file names could potentially be crafted to satisfy these conditions while attempting to influence command behavior.

next-step