lyt-product-title
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data provided by the user, including product descriptions, existing titles, keyword lists, and competitor data (ingested via the prompt in
SKILL.md). - Ingestion points: User-supplied product facts, existing titles, and external keywords provided during the initial input and diagnosis phases.
- Boundary markers: The skill uses structured templates (e.g., the '首轮输入' block) to organize information, but lacks explicit technical delimiters or escaping mechanisms to isolate user content from instructions.
- Capability inventory: The skill performs content generation, diagnostic analysis, and title optimization based on logic reasoning. It does not perform file writes, network requests, or subprocess executions.
- Sanitization: The skill includes instructions to verify keywords against a 'Fact Card' and evidence rules (
references/title-evidence.md) to filter unverified claims, providing a logical barrier against data poisoning.
Audit Metadata