The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands such as git, node, pnpm, ls, and mkdir to interact with the local repository, check environment status, and run its scanning engine.
[DYNAMIC_EXECUTION]: The skill dynamically calculates its installation path at runtime to locate and execute its internal scripts (_resolve-plugin-root.sh and scan.mjs) using the source and node commands.
[INDIRECT_PROMPT_INJECTION]: The skill scans and interprets unfamiliar codebases to identify architectural patterns. This behavior introduces a vulnerability surface where malicious instructions hidden in the analyzed source code (e.g., in comments or metadata) could potentially mislead the discovery subagent.
Ingestion points: The skill ingests codebase file contents, directory structures, and symbols via the @unwind/core scanner and subsequent file reads.
Boundary markers: The instructions do not define strict delimiters or 'ignore' instructions for the subagent when processing the potentially untrusted source data.
Capability inventory: The skill has permissions to read and write to the file system (Read, Write, Edit) and execute shell commands via Bash.
Sanitization: There is no evidence of sanitization or filtering performed on the source code content before it is processed by the LLM subagent.

uw-scan