create-context
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external data (codebase entities, types, and database tables) to propose domain terms. This creates a surface for indirect prompt injection if an attacker embeds malicious instructions within the source code files.
- Ingestion points: The Phase 1: explore step instructions the agent to read arbitrary codebase files to find raw material.
- Boundary markers: The skill implements a human-in-the-loop validation, requiring user confirmation for every term and definition before writing to the filesystem.
- Capability inventory: The skill utilizes file-read operations across the codebase and file-write operations specifically for CONTEXT.md and related context files.
- Sanitization: No explicit code-level sanitization is performed on the ingested strings; however, the iterative Q&A process acts as a functional filter for anomalous or malicious content.
Audit Metadata