research-methods

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from web sources and local files to generate research documents.\n
  • Ingestion points: External web content retrieved via WebSearch and WebFetch tools, and local codebase content accessed via Glob, Grep, and Read tools.\n
  • Boundary markers: The prompt templates for sub-agents (e.g., Web Research Agent, Community Agent) lack explicit instructions or delimiters to isolate user-provided or fetched data from the system instructions.\n
  • Capability inventory: The skill possesses the capability to read any file within the project scope and write research findings back to the filesystem using the Write tool.\n
  • Sanitization: There is no evidence of content sanitization or validation to prevent malicious instructions embedded in data from being interpreted by the LLM.\n- [EXTERNAL_DOWNLOADS]: Fetches documentation and package metadata from well-known technology sites and official registries like npmjs.com, pypi.org, and pkg.go.dev to provide accurate research comparisons. This is part of the primary intended functionality for technical research.\n- [COMMAND_EXECUTION]: The skill utilizes local file system exploration tools (Glob, Grep, Explore) to analyze project structure, dependencies, and code patterns for its research and refactoring workflows. These operations are conducted using standard agent tools and are restricted to the project context.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 05:50 AM