awslabs / awslabs/aidlc-workflows
[Bug]: Rule files reference stages that do not exist in the current workflow ("Unit Design Generation", "NFR Implementation")
- Dominant language
- TypeScript
- Stars
- 4.6k
- Forks
- 827
- Avg merge
- 1d 21h
- Merged PRs (30d)
- 165
Description
### Description
Several rule files still use stage names from an earlier workflow structure. They match nothing in the current stage set, in aidlc-state.md, or in the execution plan template.
- `construction/code-generation.md` (Prerequisites): requires "Unit Design Generation" and "NFR Implementation" to be complete; Step 1 says "Read unit design artifacts from Unit Design Generation"
- `inception/workflow-planning.md` (Step 3.4): the decision is titled "NFR Implementation" while the execution plan template in Step 7 has separate rows for NFR Requirements and NFR Design
- `common/error-handling.md`: section titles "Design Errors", "NFR Implementation Errors", and "Operations Errors" use old names, and one workaround defers test work to the placeholder Operations phase instead of Build and Test
- `common/terminology.md`: examples cite "Context Assessment" and "Requirements Assessment" stages, say the CONSTRUCTION phase contains 7 stages (it defines 6), and still teach "Unit Design Planning -> Unit Design Generation" sub-step names
### Steps to Reproduce
1. Run the AI-DLC workflow until Code Generation starts for any unit
2. The model loads construction/code-generation.md and evaluates the prerequisite "Unit Design Generation must be complete for the unit"
3. Inspect aidlc-docs/aidlc-state.md: stage checkboxes are named Functional Design, NFR Requirements, NFR Design, Infrastructure Design; no entry matches the prerequisite
### Expected Behavior
Gate prerequisites reference stage names that actually appear in aidlc-state.md, so the check is deterministic and reproducible across models.
### Actual Behavior
The model has to interpret the phantom names. Depending on the model and run, it infers the intent correctly, searches aidlc-docs for artifacts that never existed, asks the user a confusing question, or treats the prerequisite as unmet. This is a source of run-to-run variance.
### AI-DLC Version
v1 (current)
### Release / Commit
_No response_
### AI-DLC Phase
Inception (requirements, user stories, application design)
### Platform / IDE
Other
### AI Model
_No response_
### Environment
_No response_
### Additional Context
A reference fix is available on my fork: https://github.com/Emillock/aidlc-workflows/tree/rules-fixes-reference (commits ff5e9a6 and 11b6f3d).
Contributor guide
Research direction
Start by comparing construction/code-generation.md, inception/workflow-planning.md, common/error-handling.md, and common/terminology.md with aidlc-docs/aidlc-state.md and the execution plan template. Use the reference fix linked in the issue to identify the intended names and verify that all stage references, phase counts, and test-work guidance match the current workflow.
Written by the indexing model from the issue text.
Assessment
- Domain
- documentation
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Active
- Clarity
- Clearly specified
- Newbie friendliness
- 68/100