instrument-agent

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches SDKs (progress-observability, @progress/observability, Progress.Observability.Instrumentation) from official registries (npm, PyPI, NuGet). These represent legitimate vendor packages associated with Progress and Telerik.
  • [INDIRECT_PROMPT_INJECTION]: The skill architecture handles untrusted trace content through shared MCP tools, creating an attack surface. 1. Ingestion points: Data enters via get_observation_details_with_content. 2. Boundary markers: Content must be treated as data, not instructions. 3. Capability inventory: The skill performs file writes and package management but is restricted from calling MCP read tools itself. 4. Sanitization: Instructions require defanging boundary tokens (e.g., to < input >) and PII scrubbing.
  • [CREDENTIALS_UNSAFE]: The skill follows strict secret management by using environment variables (OBSERVABILITY_API_KEY) and explicitly prohibiting hardcoded keys or user elicitation. It enforces loud failures for missing keys to prevent silent telemetry loss.
  • [COMMAND_EXECUTION]: The skill utilizes standard package managers (pip, npm, dotnet) to install dependencies and run the agent for telemetry verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 06:32 AM
Security Audit — agent-trust-hub — instrument-agent