bug-clustering

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external social media data, creating an attack surface for indirect prompt injection.\n
  • Ingestion points: The skill ingests X post objects from user-supplied input files as specified in the argument-hint.\n
  • Boundary markers: No automated delimiters are used for the input data; instructions rely on the agent manually inspecting the schema and refusing unknown binary sources.\n
  • Capability inventory: The skill performs file writes to a local SQLite database and executes project-specific shell commands using the Bun runtime.\n
  • Sanitization: The skill implements redactPii() and parseCandidate() functions to mask sensitive content, although the runtime contract explicitly notes these are not exhaustive controls and treats operator review as a separate boundary.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 05:53 PM
Security Audit — agent-trust-hub — bug-clustering