scientific-critical-thinking
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command
python scripts/generate_schematic.py "your diagram description" -o figures/output.png. The filescripts/generate_schematic.pyis missing from the skill's files, resulting in an instruction to execute unverified code. - [PROMPT_INJECTION]: The schematic generation instructions create an indirect prompt injection surface (Category 8) by interpolating user-controlled data into a shell command template.
- Ingestion points: User-provided "diagram description" input described in
SKILL.md. - Boundary markers: Absent; the user input is placed directly inside double quotes within a bash command.
- Capability inventory: The skill frontmatter allows the
Bash,Write,Edit, andReadtools. - Sanitization: Absent; there are no instructions to sanitize or escape user input before it is passed to the shell.
- [PROMPT_INJECTION]: There is an inconsistency in the skill's metadata (Category 7). The YAML frontmatter lists the author as 'K-Dense Inc.', while the provided author context is 'qinyan-ai'.
Audit Metadata