software-workflow
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill includes instructions that attempt to override external behavior and mandate strict adherence to its own documentation.
- Evidence: SKILL.md states: "重要指令：你必须完整阅读本文档引用的所有文件... 规则之间有冲突时，以 workflow-rules.md 为准。"
- [COMMAND_EXECUTION]: The skill instructs the agent to perform local environment modifications using git commands.
- Evidence: workflow-rules.md directs the agent to "检查是否已有 git 仓库，若没有则执行 git init，并提交全部文件".
- [INDIRECT_PROMPT_INJECTION]: The workflow involves processing project requirements, designs, and code which are external data sources that could influence agent behavior.
- Ingestion points: The skill ingests requirement analysis from sub-modules and handles project-wide source code.
- Boundary markers: There are no explicit markers or instructions to disregard embedded directives in the ingested project files.
- Capability inventory: The skill has the capability to execute git commands and perform file operations during coding and deployment stages.
- Sanitization: No sanitization or validation of the processed data is mentioned.
Audit Metadata