cursor-workflow-from-chats
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes chat transcripts to generate persistent agent instructions, creating a vector for malicious data to influence the agent.\n
- Ingestion points: Ingests chat transcripts from project-local JSONL files as defined in SKILL.md and scripts/cursor_canvas_adapters.py.\n
- Boundary markers: Absent. The skill lacks delimiters or protective instructions to distinguish between benign chat history and potential embedded commands.\n
- Capability inventory: The skill possesses the ability to create or update project-level files, specifically 'new skill', 'skill edit', 'rule', and 'workflow doc', which can durably change agent behavior.\n
- Sanitization: The scripts/cursor_canvas_adapters.py file implements a redaction function (_redact_preference_text) for secrets and private identifiers, but does not provide validation or sanitization for natural language instructions.
Audit Metadata