verify-chain
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted markdown articles to extract technical assertions, which are then used to generate web search queries and modify local files.
- Ingestion points: The workflow begins by reading a user-provided markdown file in Phase 1 of the process defined in SKILL.md.
- Boundary markers: The skill utilizes independent system prompts for each role (Critic, Verifier, Repairer) to enforce context isolation and minimize the influence of user-provided content on the agent's core behavior, as seen in prompts/verifier.md and prompts/repairer.md.
- Capability inventory: The skill is granted WebSearch, WebFetch, and Write capabilities to verify claims and produce modified articles and audit logs.
- Sanitization: The skill implements a strict regex validation (^[a-z0-9][a-z0-9-]{0,39}$) for the article slug used in file path construction, preventing path traversal and shell injection risks during the generation of the verification report in Phase 4.
Audit Metadata