The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses shell commands and environment variables to compute the path to its own scripts before executing them. Specifically, it searches the user's plugin cache directory to define the UNWIND_PLUGIN_ROOT variable.
Evidence: UNWIND_PLUGIN_ROOT="$(ls -dt "$HOME"/.claude/plugins/cache/*/unwind/*/ 2>/dev/null | head -1)" and source "${UNWIND_PLUGIN_ROOT%/}/skills/scripts/_resolve-plugin-root.sh".
[INDIRECT_PROMPT_INJECTION]: The skill processes source code files from the local repository to identify structural changes and updates analysis documentation based on that content. This content is then passed to other agent specialists as context.
Ingestion points: Codebase files within the current working directory ($(pwd)), and cached JSON metadata files (docs/unwind/.cache/meta.json).
Boundary markers: No explicit boundary markers or delimiters are used when passing ingested code content to other tools or specialists.
Capability inventory: The skill can execute shell commands (git, node, pnpm), write to and edit files in the docs/unwind/ directory, and dispatch tasks to other agents.
Sanitization: No sanitization of ingested code content is performed before it is used for documentation generation or as context for other specialists.
[COMMAND_EXECUTION]: The skill executes multiple bash commands and Node.js scripts to perform its analysis tasks.
Evidence: Use of node to run detect-changes.mjs, scan.mjs, seed-layers.mjs, verify-coverage.mjs, and build-graph.mjs.

uw-refresh