The Agent Skills Directory

[CREDENTIALS_UNSAFE] (SAFE): The skill identifies hardcoded credentials as a critical risk in its 'Sharp Edges' documentation and provides regex validations to detect them in user code. No actual secrets are leaked; examples use placeholders.
[EXTERNAL_DOWNLOADS] (SAFE): Reference patterns utilize standard, trusted Docker images (node:20-slim, redis:7-alpine). No untrusted remote downloads or piped bash scripts are present.
[COMMAND_EXECUTION] (SAFE): The skill describes Docker commands and shell signals (SIGTERM/SIGINT) strictly for orchestration and lifecycle management purposes. No arbitrary or malicious command execution is present.
[DATA_EXFILTRATION] (SAFE): Network operations are limited to health checks (localhost) and standard database/cache connections. No sensitive file access or unauthorized network sends are performed.
[PROMPT_INJECTION] (SAFE): No instructions attempt to bypass agent safety filters, override system prompts, or extract internal instructions.

mcp-deployment