The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes terminal commands using project file paths and user arguments, which may not be properly sanitized for shell metacharacters.
Evidence: Step 7 in SKILL.md instructs the agent to execute npx playwright test <converted-file> --reporter=list. The <converted-file> variable is derived from file discovery or the $ARGUMENTS input, creating a vector for command injection if a filename contains shell-active characters (e.g., ;, &, |).
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted code from external test files and project configurations, then performs actions based on that content.
Ingestion points: Test files in the cypress/ directory, Selenium .py files, and dependencies listed in package.json (detected in Step 1).
Boundary markers: Absent. There are no instructions to the agent to treat the contents of the migrated files as data only or to ignore embedded instructions.
Capability inventory: The agent has the ability to write files (Step 4), delete files (Step 8), and execute terminal commands (npx, npm).
Sanitization: Absent. The skill does not include validation steps to ensure that the source test files do not contain malicious natural language instructions designed to hijack the agent's migration logic.

migrate