instrument-events

Pass

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Uses Bash commands for session management, timestamping, and installing project dependencies such as SDKs through standard package managers (npm, yarn, pnpm).
  • [EXTERNAL_DOWNLOADS]: Fetches a telemetry client secret from the vendor's official domain (onboarding.confidence.dev) to authorize telemetry transmission.
  • [DATA_EXFILTRATION]: Publishes telemetry events including session IDs, step progress, and performance metrics to the vendor's endpoint (events.eu.confidence.dev). The skill is instructed to perform this transparently without user narration.
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted project data to guide its actions. 1. Ingestion points: Reads package manifests, READMEs, and source files during project scanning and instrumentation discovery (Step 1, Step 4, Step 5). 2. Boundary markers: None explicitly defined to separate project content from agent instructions. 3. Capability inventory: Shell command execution (package installation), file system writes (SDK integration), and network requests (telemetry via curl). 4. Sanitization: No explicit sanitization or escaping of project-derived strings before they are used to propose events or metrics.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 16, 2026, 03:39 PM