skf-export-skill

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a workflow to inject content from external skill artifacts into project-level instruction files like .cursorrules and CLAUDE.md, which define agent behavior and context. This creates a surface for indirect prompt injection.
  • Ingestion points: Skill artifacts are loaded from metadata.json and SKILL.md in steps-c/step-01-load-skill.md and steps-c/step-03-generate-snippet.md.
  • Boundary markers: The skill uses <!-- SKF:BEGIN --> markers to wrap the managed section in target files, but it does not provide internal sanitization or delimiters for individual snippets to prevent malicious instructions within a skill from overriding the agent's constraints.
  • Capability inventory: The skill performs file-write operations on sensitive agent instruction files (CLAUDE.md, .cursorrules, AGENTS.md) in steps-c/step-04-update-context.md.
  • Sanitization: No validation or sanitization of the content extracted from skill artifacts is performed before injection into the project configuration files.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 05:05 PM
Security Audit — agent-trust-hub — skf-export-skill