intercom-common-errors

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXFILTRATION]: The skill utilizes the INTERCOM_ACCESS_TOKEN environment variable to authenticate requests to official Intercom API endpoints (api.intercom.io). Network communication is appropriately scoped to well-known service domains for diagnostic purposes and does not attempt to exfiltrate sensitive data to unauthorized third parties.
  • [EXTERNAL_DOWNLOADS]: The instructions recommend installing the official intercom-client SDK via NPM. This is a standard, versionable, and verifiable dependency for the targeted service and does not represent a remote code execution risk.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from logs and source code to debug API errors. While this involves ingesting untrusted content, the risk is assessed as safe because the operation is limited to identifying machine-readable error codes and does not involve the execution of data-derived commands or dangerous logic.
  • Ingestion points: Project logs and source code (SKILL.md).
  • Boundary markers: Boundary markers are not explicitly defined in the provided snippets, as the skill relies on standard agent behavior for data ingestion.
  • Capability inventory: File reading (Read), code searching (Grep), and authenticated network requests via shell commands (curl) (SKILL.md).
  • Sanitization: Sanitization is not required for the identified diagnostic use case, as the outputs are intended for human-in-the-loop debugging.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:45 AM
Security Audit — agent-trust-hub — intercom-common-errors