The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several shell commands for file inspection, text manipulation, and verification.
Uses git commands (status, diff) to identify changes.
Employs perl -pi -e for automated search-and-replace normalization of em-dash characters.
Executes project-level verification tools via npx (tsc, eslint, prettier) and npm test to ensure code quality and behavioral parity.
[INDIRECT_PROMPT_INJECTION]: The skill includes a vulnerability surface for indirect prompt injection by reading external data that could influence the agent's behavior.
Ingestion points: The workflow specifically instructs the agent to read context from /tmp/<dirname>/code-review.md in step 2.
Boundary markers: Absent. The instructions do not define delimiters or provide specific warnings to the agent to treat the contents of the review file as untrusted data.
Capability inventory: The agent has the capability to write to the filesystem (via perl) and execute arbitrary code defined in the project's test scripts (via npm test).
Sanitization: No validation or sanitization of the review file content is performed before the agent acts upon the findings listed within it.

typescript-refactor