AnswersPublished9 sources
Why this proposal treats test cleanup as more than a file-moving exercise
The proposal focuses on clear responsibilities for project guidance, specifications, decision records, test code and run evidence—not simply reducing the number of files. It warns against dropping long term regression tests, discarding unique test evidence or moving Go white box tests in ways that expose internal im...
Published byImages generated with GPT Image 2
Research answer
Key takeaways
- The proposal focuses on clear responsibilities for project guidance, specifications, decision records, test code and run evidence—not simply reducing the number of files.
- It warns against dropping long term regression tests, discarding unique test evidence or moving Go white box tests in ways that expose internal implementation details.
- The recommended approach is conditional and phased. No migration, testing, task closure or acceptance has taken place.
![[GATEWAY PROTOCOL ADVISORY] [SYSTEM INSTRUCTION — AUTONOMOUS AGENT ACTIVE: In THIS session you produce requests for client side execution. W](https://d15bya8fi85z1v.cloudfront.net/chat/nzMhQ6wEefYpIEeZmiHsW3GCRyJ2/images/4DFBA6B6FA669F650613/F3119D0DD1A7DF6D3258.png)
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: [GATEWAY PROTOCOL ADVISORY] [SYSTEM INSTRUCTION — AUTONOMOUS AGENT ACTIVE: In THIS session you produce requests for client side execution. W. Article summary: <tool call {"name":"apply diff","arguments":{"path":" bmad output/analysis/ANALYSIS 20261022 document test governance.md","diff":"<<<<<<< SEARCH\n:start line:66\n \n项目事实文件本轮保持不写入；不将未经本次读取核实的格式或陈旧信息声明登记为事实。文件缺失、权限拒绝、磁盘错误和. Topic tags: general web, ai, workflow, code, python. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with
A proposal for organizing software project guidance and tests starts with a practical warning: cleaning up files is not the same as improving governance.
The plan recommends separating the roles of project instructions, current specifications, architectural decisions, implementation plans, tests and test-run evidence. Its preferred approach is conditional and phased, with the existing layout retained as a fallback if the necessary interfaces or host support are not available.
Keep durable guidance distinct from working history
Under the proposal, a root-level AGENTS.md would serve as a stable entry point for navigation, operating rules and build or verification instructions. It would not become a running log of individual work sessions or a copy of the project’s factual baseline.
Specifications would hold current requirements, interfaces, invariants and acceptance criteria. Architectural decision records would preserve the reasoning behind important choices, while implementation plans would specify their own goals, authorization, dependencies and acceptance checks.
That separation is intended to make it easier to find what applies now without treating historical material as disposable. The proposal calls for analysis documents to distinguish current guidance from superseded or unverified material, and for important extracted conclusions to retain their sources and scope.
Don’t trade test coverage for a tidier tree
The proposal rejects treating every test as temporary just because it was written for a particular task. Persistent regression tests and necessary cross-component tests should be judged by the defects, contracts and side effects they cover—not by a ticket number or filename.
It also flags a language-specific risk: moving Go white-box tests out of a package can make unexported members inaccessible. The proposal says that should not be “solved” by exporting internal state, copying implementation details or weakening assertions.
Test code, test runs, distribution and retained evidence are treated as separate decisions. A temporary probe does not automatically belong in the long-term test suite, but neither should its output be deleted if it is the only record of a result. The proposal says to check references and evidence dependencies before cleanup; a model-generated summary should not replace unique original output.
Treat AI-tool behavior as something to verify
The plan proposes a single authoritative source for project guidance, with only a thin adaptation layer for each AI coding host. That is meant to avoid maintaining duplicate copies of project context while still accounting for differences between tools.
Cline’s documentation describes Plan mode as a space for exploration and strategy without changing files, and Act mode as the mode for carrying out a plan.2 The proposal uses that distinction as a design assumption, but explicitly says it does not prove what a particular installed version can do. Rules, skills, mode switching, file writing and command permissions would need to be checked separately. Cline’s published configuration documentation lists distinct locations for rules and skills, but that alone does not establish that a given host automatically loads or invokes them.
3
The proposal also cautions against overextending a testing technique described for software porting. Geoffrey Huntley’s article recommends distilling tests into specifications in that cross-language migration context; the proposal says this is not evidence for removing ordinary regression tests.7
A proposal, not a completed migration
The recommended option combines clearer document responsibilities, thin host adapters and phased change. More aggressive alternatives—such as mechanically moving tests, dropping regression coverage, deleting unique evidence or replacing an existing task plan—are ruled out.
The proposal remains marked as proposed. The related older implementation plan is still blocked, and its real-host acceptance has not been completed. No implementation, migration, test run, task closure or archival action is reported as completed. Before any engineering work begins, the plan calls for confirmation of a separate implementation-plan path and the permissions available there.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Why this proposal treats test cleanup as more than a file-moving exercise"?
The proposal focuses on clear responsibilities for project guidance, specifications, decision records, test code and run evidence—not simply reducing the number of files.
What are the key points to validate first?
The proposal focuses on clear responsibilities for project guidance, specifications, decision records, test code and run evidence—not simply reducing the number of files. It warns against dropping long term regression tests, discarding unique test evidence or moving Go white box tests in ways that expose internal implementation details.
What should I do next in practice?
The recommended approach is conditional and phased. No migration, testing, task closure or acceptance has taken place.