pitch-deck-visuals
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
infshCLI to run applications such asinfsh/html-to-image,infsh/python-executor, andfalai/flux-dev-lorato generate visual content. This involves executing system commands and Python scripts provided within the skill. - [EXTERNAL_DOWNLOADS]: The documentation references and suggests installing additional skills from the
inference-shorganization using thenpx skills addcommand. - [PROMPT_INJECTION]: The skill uses templates that involve executing code or rendering HTML based on inputs. This presents a surface for indirect prompt injection if an agent populates these templates with untrusted user data without sanitization.
- Ingestion points: Markdown code blocks in
SKILL.mddesigned for tool execution (e.g.,--inputJSON strings). - Boundary markers: None identified.
- Capability inventory: Shell command execution via
infsh, Python script execution (viapython-executor), and local file creation (e.g.,market-size.png). - Sanitization: No explicit validation or escaping of user-provided content is defined for the command payloads.
Audit Metadata