deep-research-query
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted user input for research topics and interpolates it directly into generated JSON queries and Markdown briefs without sanitization or boundary markers.
- Ingestion points: User-provided topic input collected via the
AskUserQuestiontool in SKILL.md (Phase 1 and Phase 2). - Boundary markers: None present; the user input is treated as a direct parameter for the generated output.
- Capability inventory: File system writing (Phase 4) and passing data to the
deep-research-mainskill. - Sanitization: None observed for user-provided strings.
- [DATA_EXFILTRATION]: The skill uses unvalidated user input (
{topic}) to construct the file path where research queries are saved:RESEARCH/queries/{topic}_{timestamp}.json. This provides a surface for path traversal if the user provides a topic string containing directory navigation sequences (e.g.,../), potentially allowing files to be written outside the intended directory.
Audit Metadata