research-coordinator

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface where untrusted data from external files can influence agent behavior.
  • Ingestion points: The skill reads content from research/**/*-subtasks.json and research/<slug>/<slug>-research-plan.md using the file system.
  • Boundary markers: Triple quotes (""") are used to delimit the subtask description in the sub-agent prompt. However, the {topic} and {full_text_of_research_plan} variables are interpolated directly into the system instructions without strong delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill has the ability to read and write files within the research/ directory and spawn multiple independent sub-agents with custom prompt templates.
  • Sanitization: There is no evidence of sanitization, filtering, or validation performed on the text extracted from the JSON or Markdown files before it is used to construct the sub-agent prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 01:28 PM