clade-model-inference
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides legitimate instructional content for using the Anthropic Messages API with TypeScript and Python.
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates processing data (text and images) that is subsequently sent to an LLM. While this is an inherent attack surface for indirect prompt injection, the skill does not introduce any specific vulnerabilities or unsafe interpolation patterns.
- Ingestion points: Ingests user messages and image data in
SKILL.md(Steps 1, 2, and 3). - Boundary markers: None present in the code examples, which is typical for simplified documentation snippets.
- Capability inventory: Uses file read capabilities (
fs.readFileSyncforscreenshot.png) and network capabilities (Anthropic API calls) as described inSKILL.md. - Sanitization: None present in the snippets; the skill relies on the underlying LLM's guardrails.
- [DATA_EXPOSURE]: The vision example includes a file read operation (
fs.readFileSync('screenshot.png')). This is documented as a functional example for multimodal analysis and does not target sensitive directories or credentials.
Audit Metadata