agentic-rules-writer

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by scanning and incorporating data (names and descriptions) from other installed skills and agents into the generated system rules. If a user has a malicious skill installed, its description could influence the final rule file and agent behavior.
  • Ingestion points: Phase 4 scans SKILL.md and agent configuration files in multiple directories, including the user's home directory.
  • Boundary markers: The generated rule file does not appear to use strict delimiters or 'ignore embedded instructions' warnings for the data extracted from external metadata.
  • Capability inventory: The skill performs file system reads and writes persistent configuration files that direct agent behavior.
  • Sanitization: The skill does not perform explicit sanitization or validation of the metadata extracted from external files before interpolation.
  • [PROMPT_INJECTION]: The skill implements a 'Persona' feature that introduces role-playing instructions into the agent's system prompt. This is mitigated by a mandatory safety constraint included in the generated output that prioritizes technical accuracy and precision over the persona character.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 01:18 PM