browser-extension-best-practices
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze and modify external "target projects" based on a reference implementation. This creates a surface for indirect prompt injection if the target project being analyzed contains malicious instructions designed to influence the agent's behavior.
- Ingestion points: The agent is instructed to read source code, configuration files (e.g.,
package.json,wxt.config.ts), and documentation from the target project. - Boundary markers: The skill includes explicit instructions to prioritize the target project's existing constraints and use only relevant parts of the reference design ("only采用与目标项目约束匹配的部分").
- Capability inventory: The agent utilizing this skill likely possesses file system read/write capabilities and shell command execution to perform refactoring and quality checks.
- Sanitization: The skill mandates a decision-making process (keep, adopt, refactor, defer) and requires explicit user approval before any project modifications are applied.
Audit Metadata