The Agent Skills Directory

[SAFE]: The skill implements comprehensive redaction logic for environment variables. It instructs the agent to treat all values within .env files as sensitive and forbids displaying, storing, or repeating them, ensuring only the presence of keys is reported.
[SAFE]: Diagnostic workflows for port conflicts and service management are designed with a principle of least privilege. The skill mandates explicit user approval and verification of process IDs before any termination commands are suggested or executed.
[PROMPT_INJECTION]: The skill processes untrusted local data, creating a potential surface for indirect prompt injection. 1. Ingestion points: The agent reads project metadata and configuration files such as package.json, requirements.txt, and .env.example (SKILL.md, references/checks.md). 2. Boundary markers: Specific delimiters to isolate untrusted file content from system instructions are not defined. 3. Capability inventory: The skill can suggest or execute shell commands for dependency installation (npm install), environment setup (python -m venv), and process management (kill). 4. Sanitization: Risk is mitigated by strict value redaction for secrets and a requirement for explicit human-in-the-loop authorization before performing any remediation steps.

env-doctor-free