The Agent Skills Directory

[SAFE]: The skill provides standard guidance for using npm commands such as npm ci and npm install for reproducible workflows and lockfile discipline.
[PROMPT_INJECTION]: The identified 'ignore previous instructions' string in references/evals.yaml is part of a security test suite used to evaluate the agent's resistance to adversarial prompts, not a malicious instruction.
[DATA_EXFILTRATION]: The instructions include a clear constraint to redact secrets, tokens, and credentials from outputs, which is a significant security control to prevent data exposure.

npm-workflow-discipline