deep-research-report

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from web research, codebase exploration, and external data sources (databases, APIs), creating a surface for indirect prompt injection. 1. Ingestion points: Research tool (web research), codebase exploration, and data analysis in SKILL.md Section 2. 2. Boundary markers: Includes instructions to verify claims against primary sources and implementation of an adversarial review sub-agent to challenge results. 3. Capability inventory: The agent can write files, execute shell commands, and access sensitive data sources (if granted by the user). 4. Sanitization: No explicit instruction for sanitizing external content interpolated into reports.
  • [DYNAMIC_EXECUTION]: The skill generates and executes code at runtime to create data visualizations. It instructs the agent to copy a template script (scripts/charts.py), modify its execution block with research data, and run it to produce SVG artifacts.
  • [COMMAND_EXECUTION]: The skill executes external commands to build and serve the report site using uvx zensical and utilizes standard utilities like qlmanage, rsvg-convert, or ImageMagick for visual chart verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 03:51 PM