The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to run npx ai-devkit@latest memory store to record failure patterns. This fetches the ai-devkit package from the npm registry (a well-known service) at runtime, though the use of the @latest tag prevents version pinning.
[COMMAND_EXECUTION]: The skill's core function involves identifying and executing shell commands to verify task outcomes. This is a deliberate design for its 'Gate Function', allowing the agent to run arbitrary code to satisfy its verification rules.
[PROMPT_INJECTION]: The skill interprets terminal output to verify task completion, which presents a surface for indirect prompt injection (Category 8).
Ingestion points: Terminal output from commands executed in the Run and Read steps.
Boundary markers: There are no instructions to use boundary markers or delimiters when reading command output to distinguish it from instructions.
Capability inventory: The agent can execute terminal commands and invoke the ai-devkit CLI.
Sanitization: No sanitization or validation logic is defined for the command output before the agent evaluates its contents.

verify