api-mongo-agg-facet-bypass

Warn

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill provides structured methodologies and JSON payload templates specifically designed to exfiltrate sensitive secrets, including invite tokens, credentials, password hashes, session documents, and API keys from sibling MongoDB collections.
  • [COMMAND_EXECUTION]: The instructions describe a technique to bypass server-side allowlists by nesting restricted MongoDB aggregation stages (such as $unionWith and $lookup) inside a permitted $facet stage, effectively executing unauthorized queries against the database.
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies a vulnerability surface where an API processes untrusted, user-supplied pipeline data, which can be manipulated to bypass security constraints.
  • Ingestion points: The pipeline parameter in API search endpoints (described in SKILL.md).
  • Boundary markers: The analysis notes that allowlists typically only check top-level keys, failing to secure nested sub-pipelines.
  • Capability inventory: The skill enables the agent to perform cross-collection reads and brute-force collection names to locate sensitive data.
  • Sanitization: The skill highlights the lack of recursive validation in the target application's stage allowlist.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 20, 2026, 11:02 PM
Security Audit — agent-trust-hub — api-mongo-agg-facet-bypass