skills/simota/agent-skills/mend/Gen Agent Trust Hub

mend

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is an operationally focused SRE tool with comprehensive security documentation. It implements a 4-tier risk model (T1-T4) that classifies actions based on blast radius, reversibility, and data sensitivity, ensuring that potentially dangerous operations (like DNS or config changes) require explicit human approval.- [PROMPT_INJECTION]: The skill is designed to handle external data inputs from monitoring systems, which is a potential surface for indirect prompt injection. However, the skill explicitly addresses this via a dedicated defense protocol.
  • Ingestion points: Processes data from TRIAGE_TO_MEND_HANDOFF, BEACON_TO_MEND_HANDOFF, and _AGENT_CONTEXT which include incident diagnoses and alert details.
  • Boundary markers: Utilizes structured format validation (JSON/YAML) and enforces the separation of user-generated content from metric signals.
  • Capability inventory: Includes the ability to restart pods, scale services, rollback configurations, and manage circuit breakers.
  • Sanitization: Implements a three-step validation pipeline that includes stripping user content from telemetry and matching against canonical error codes rather than free-text descriptions.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 04:22 PM