The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill includes a Python code snippet with a hardcoded bearer token pattern 'sk-dev-proxy-123'. While presented as a placeholder for a local development proxy, it matches high-entropy credential patterns.
Evidence: headers={"Authorization": "Bearer sk-dev-proxy-123"} in SKILL.md.
[COMMAND_EXECUTION]: The instructions direct the agent to execute system management commands to control local daemon services.
Evidence: systemctl --user start embry-memory in SKILL.md.
[PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface (Category 8) by defining workflows where an agent processes untrusted data and executes automated repair orchestration.
Ingestion points: 'WebGPT responses', 'source text', and 'manifests' (SKILL.md).
Boundary markers: No requirement for data delimiters or 'ignore' instructions for embedded content was identified.
Capability inventory: The agent is authorized to perform local network operations via httpx, execute system commands, and run automated repair logic.
Sanitization: The framework lacks specific rules for sanitizing or validating external content before it influences automated decisions or repair actions.

best-practices-agent