constructive-auth

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill provides instructions and SDK examples for processing untrusted external data, specifically OAuth profile information (details object) and user-supplied identity identifiers. This data is interpolated into core authentication flows and identity linking logic.
  • Ingestion points: In references/identity-linking.md, OAuth profile details are ingested via the signInIdentity and linkIdentity mutations.
  • Boundary markers: The provided code snippets do not demonstrate explicit delimitation, schema validation, or sanitization of the structured details JSON object before it is passed to the database functions.
  • Capability inventory: The skill facilitates powerful database mutations, including session creation, security setting updates (app_settings_auth), and device trust management.
  • Sanitization: While the documentation correctly highlights the use of Bcrypt for password credentials, it does not explicitly instruct on the sanitization of metadata from third-party identity providers, which could be exploited if that data is later used in sensitive contexts.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 01:06 PM
Security Audit — agent-trust-hub — constructive-auth