ai-rag-pipeline

Pass

Audited by Gen Agent Trust Hub on Apr 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of the infsh command-line tool via bash to run AI models and search engines.
  • Evidence: Frequent use of infsh app run commands throughout the SKILL.md examples.
  • Evidence: Use of shell variable interpolation (e.g., "$SEARCH_RESULT") to pass data from one command to another, which can lead to command injection or JSON parsing errors if the variable content contains shell metacharacters or unbalanced quotes.
  • [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from web searches and interpolates it directly into LLM prompts, creating a surface for indirect prompt injection.
  • Ingestion points: Search results from Tavily and Exa are stored in variables and then inserted into subsequent model prompts in SKILL.md.
  • Boundary markers: Absent. There are no delimiters or specific instructions for the AI to ignore instructions embedded within the retrieved context.
  • Capability inventory: The agent can execute shell commands via infsh and access external APIs.
  • Sanitization: Absent. External content is used raw as retrieved from the search tools.
  • [EXTERNAL_DOWNLOADS]: The skill points users toward external resources for installation and dependency management.
  • Evidence: Mentions an installation guide for the CLI hosted at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
  • Evidence: Encourages the use of npx skills add to install additional components from the vendor's repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 23, 2026, 12:24 AM