tw-research-grounded-dialogue
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a research-oriented dialogue system using simulated personas. The tool usage is constrained to typical research operations like reading local reference files and performing web searches for grounding. No evidence of obfuscation, privilege escalation, or unauthorized data access was found.\n- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection as it processes data from external web searches and user-supplied research drafts.\n
- Ingestion points: File
SKILL.md(Step 3: WebSearch results) andreferences/dialogue-formats.md(Stage C: Writing - User-provided research sections).\n
- Boundary markers: There are no specific instructions or delimiters defined to isolate external data or to instruct the agent to ignore instructions embedded within the search results.\n
- Capability inventory: The skill is granted access to
Bash,Read,Write, andWebSearchtools, allowing for potential tool abuse if the agent obeys instructions from untrusted external sources.\n - Sanitization: The instructions do not specify any validation or filtering logic for external content retrieved via tools.
Audit Metadata