ce-brainstorm

Pass

Audited by Gen Agent Trust Hub on Jul 7, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-supplied input through the #$ARGUMENTS variable and reads existing brainstorm documents. This constitutes an indirect prompt injection surface.
  • Ingestion points: User-provided feature descriptions and existing files in docs/brainstorms/ are read into the context.
  • Boundary markers: Input is wrapped in <feature_description> tags, but the instructions do not explicitly direct the model to ignore malicious commands embedded in this data.
  • Capability inventory: The skill can write to the local file system, call other engineering tools (ce-plan, ce-work), and interact with external services (Slack, Every's Proof).
  • Sanitization: No validation or sanitization of input data is described in the logic.
  • [DATA_EXFILTRATION]: The skill provides documented functionality to move data to external platforms.
  • It integrates with the ce-proof tool to upload requirements to an external web editor (Every's Proof).
  • It uses the ce-slack-researcher tool to access internal organizational context. These are features of the vendor's integrated product suite.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 7, 2026, 08:06 AM
Security Audit — agent-trust-hub — ce-brainstorm