down-skilling
Pass
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or sensitive data access was detected. The skill is composed entirely of Markdown documentation and prompt templates providing logic for AI reasoning distillation.
- [INDIRECT_PROMPT_INJECTION]: The skill defines templates for processing potentially untrusted external data (e.g., resumes, emails, and customer feedback). It mitigates risks by implementing security best practices such as structured delimiters and strict context-grounding rules.
- Ingestion points: Prompt templates in the
examples/directory (e.g.,data-extraction.md,resume-screening.md) accept untrusted user content via placeholders like{{comments}}and{{resumes}}. - Boundary markers: The skill prescribes the use of XML-based tagging (e.g.,
<context>,<rules>,<examples>) to isolate untrusted data from the agent's instructions. - Capability inventory: None. No code execution, file system access, or network capabilities are defined or utilized within the skill files.
- Sanitization: Prescribes strict 'use ONLY information from context' rules and mandates specific scripted responses for missing or insufficient data.
Audit Metadata