optimize-agents-md
Pass
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core design of reading and auditing existing instruction files.
- Ingestion points: In
SKILL.md(Stage 1), the agent is instructed to readAGENTS.md,CLAUDE.md,.claude/,.cursor/rules/,.cursorrules, and other project files. - Boundary markers: The instructions do not define delimiters or specific "ignore embedded instructions" warnings when reading these files.
- Capability inventory: The skill has the capability to write and modify files (
AGENTS.md,CLAUDE.md) as part of its Drafting (Stage 5) and Wiring (Stage 6) phases. - Sanitization: There is no mention of sanitizing, escaping, or validating the content extracted from these files before using it to generate new instructions.
Audit Metadata