scientific-critical-thinking
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONINGCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and critically analyze external scientific claims and research papers, which constitutes a primary ingestion point for untrusted data. There are no explicit instructions, delimiters, or boundary markers provided to the agent to isolate these inputs or ignore potential instructions embedded within the text. The agent's available tools (Read, Write, Edit) and its capability to execute local scripts provide an attack surface if the analyzed content contains adversarial prompts. No specific input sanitization or validation protocols are defined for the analyzed documents.
- [METADATA_POISONING]: The skill's YAML frontmatter in 'SKILL.md' identifies the author as 'K-Dense Inc.', which differs from the provided deployment context of 'eturkes', representing an inconsistency in metadata.
- [DYNAMIC_EXECUTION]: The skill instructions include a specific command to execute a local Python script, 'scripts/generate_schematic.py', for generating diagrams via the OpenRouter API. This involves the runtime execution of a local script based on natural language input generated by the agent.
- [COMMAND_EXECUTION]: The skill explicitly directs the agent to use the shell to execute a Python-based CLI tool for schematic generation, a capability that, while intended for visual aids, allows for system-level interaction.
Audit Metadata