nps-response-handler

Fail

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: CRITICALPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection as it processes untrusted user-generated content (NPS comments) without security boundaries or sanitization. An attacker could embed instructions in a survey comment to manipulate the agent's drafted replies or proposed next steps.
  • Ingestion points: Untrusted NPS scores and comments ingested via SKILL.md.
  • Boundary markers: Absent; the workflow interpolates user comments directly into the context used for drafting replies without delimiters.
  • Capability inventory: The agent can draft outgoing email replies and propose task assignments/owners.
  • Sanitization: No input validation or filtering is implemented for the NPS comment field.
  • [PROMPT_INJECTION]: There is a discrepancy in the skill's identity; the author is listed as Builder.org in the metadata, which is flagged by automated scanners as a malicious domain, while the developer context identifies the author as OnStartups. This manifest inconsistency is a common pattern in metadata poisoning or impersonation attempts.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Aug 4, 2026, 04:37 PM
Security Audit — agent-trust-hub — nps-response-handler