The Agent Skills Directory

[SAFE]: The skill follows established security best practices for application instrumentation.
[CREDENTIALS_UNSAFE]: The skill explicitly warns against hardcoding credentials. It instructs the agent to retrieve tokens via a secure MCP tool (projects-get) or prompt the user, and to store them in .env files using standard environment variable conventions.
[EXTERNAL_DOWNLOADS]: The skill references the installation of official OpenTelemetry and PostHog packages. These are well-known, trusted libraries from reputable sources (OpenTelemetry and PostHog).
[INDIRECT_PROMPT_INJECTION]: The skill includes instructions to read and modify application source code, which technically constitutes an attack surface. However, it mitigates risk by:
Ingestion points: Analyzing local project files (package.json, requirements.txt, etc.) and existing log code.
Boundary markers: Explicitly requiring the agent to read files before writing and to make minimal, targeted additions without replacing existing logic.
Capability inventory: Using standard package managers and file-write operations to add instrumentation.
Sanitization: Providing detailed best practices for structured logging and automatic PII scrubbing to prevent accidental sensitive data exposure.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard package managers (npm, pip, go get, etc.) to install necessary dependencies, which is appropriate for the stated purpose of log instrumentation.

instrument-logs