kf-g-skill-review-modification-sample-artifacts

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill defines a process where external, untrusted input data is interpolated into prompt templates for sub-agents, creating a surface for potential injection.
  • Ingestion points: The <INPUT_DATA> placeholder in references/fixed-prompt-template.md allows arbitrary content to be injected into the implementation-worker's prompt.
  • Boundary markers: The template uses Markdown code blocks (```) to encapsulate input data, which provides limited protection against structured data escape.
  • Capability inventory: The skill has the capability to spawn sub-agents (Worker A/B), read skill definition files, and write generated artifacts to the local directory artifacts/skill-modification-review/.
  • Sanitization: There is no explicit requirement for the orchestrator to sanitize, filter, or validate the <INPUT_DATA> before it is passed to the sub-agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 09:44 PM
Security Audit — agent-trust-hub — kf-g-skill-review-modification-sample-artifacts