papercodex-agent-workspace
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection as it directs the agent to read and follow instructions from external files.
- Ingestion points: The agent reads 'agent_instructions.md', 'prompt_contract.md', and 'workspace_manifest.json' from the local workspace.
- Boundary markers: Absent. The instructions do not define delimiters or provide guidance to ignore potentially malicious content within these files.
- Capability inventory: The skill allows the agent to read and write files in the workspace and execute app state changes via MCP tools (e.g., tags, notes, library mutations).
- Sanitization: Absent. There is no instruction to validate or sanitize the content of the ingested markdown or JSON files before processing.
Audit Metadata