The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze project source code, terminal outputs, and test results, which constitutes an untrusted data ingestion surface where malicious content in a repository could attempt to influence the agent.
Ingestion points: PROJECT-VERIFICATION.md, CLI-HARNESS.md, and INTEGRATION.md describe reading repository files, startup logs, and terminal transcripts as part of the testing workflow.
Boundary markers: No specific delimiters or instruction-isolation markers are mandated for the data being analyzed.
Capability inventory: The skill leverages terminal harnesses (tmux, PTY), test runners (vitest), and containerized infrastructure management (testcontainers).
Sanitization: PROJECT-VERIFICATION.md includes explicit instructions to "Keep secrets out of definitions and artifacts."
[COMMAND_EXECUTION]: The skill facilitates the execution of shell commands and the management of terminal environments to perform project verification.
Evidence: CLI-HARNESS.md provides instructions for driving interactive terminal sessions with tmux and PTY drivers. INTEGRATION.md includes TypeScript code for orchestrating Docker containers using the @testcontainers/postgresql library.

testing