prompt-engineering

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill provides numerous templates and patterns that interpolate untrusted external data (e.g., {user_input}, {target_input}, {customer_inquiry}) directly into instructions. This creates a potential surface for indirect prompt injection if the user-provided content contains malicious directives intended to override the agent's logic.
  • Ingestion points: Placeholders for untrusted data are present in SKILL.md, references/few-shot-patterns.md, and references/template-systems.md.
  • Boundary markers: While the templates use structural headers (e.g., ## Input), they generally lack strict delimiters or explicit instructions to the model to ignore potential directives within the interpolated data.
  • Capability inventory: The skill metadata specifies access to Bash, Write, and Edit tools.
  • Sanitization: The skill identifies "Input Validation" and "Injection Prevention" as best practices in references/template-systems.md, although runtime implementation of these filters is not present in the provided examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 02:30 PM