agent-lifecycle

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by instructing the agent to ingest and display telemetry from external child processes.
  • Ingestion points: Lifecycle events are ingested via JSONL streams and authoritative snapshots as defined in SKILL.md and adapter-contract.md.
  • Boundary markers: The instructions lack specific delimiters or directives to ignore instructions that might be embedded within the 'activity' or 'details' data fields.
  • Capability inventory: The agent is directed to display child status, activity strings, and transcript references to the user (SKILL.md, Usage Examples).
  • Sanitization: While the procedure includes redacting secrets, it does not provide methods for sanitizing or escaping natural language instructions contained in the processed data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 10:03 PM
Security Audit — agent-trust-hub — agent-lifecycle