docs(plans): design milestone 7 - dogfooding strategy and capability-gap decision framework - #38
Conversation
…gap decision framework Turns Milestone 7's "record missing capabilities" scope into a concrete process: migration-driven evidence (not synthetic spikes) against the real ncipollina/cosmere-tracker AutoFixture test kit, a four-question rubric per candidate gap, and a decision recorded either as a new Proposed ADR (roadmap candidate) or a dated Amendment to the governing existing ADR (intentional design difference) - so a gap can end in "no change" without becoming invisible. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ca9e4a2360
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| The migrated `cosmere-tracker` test suites passing under Compono, in that | ||
| repo, is itself the primary verification — there is no new automated test | ||
| added to the `compono` repo by this plan (it produces documentation and | ||
| decision records, not product code). If a gap's outcome is a "roadmap |
There was a problem hiding this comment.
If dogfooding exposes confusing diagnostics or an unnecessary abstraction, this test plan explicitly expects no product-code work, and none of its phases includes a task to refine or remove either; however, docs/mvp.md lines 421-422 still define both as Milestone 7 scope. The plan can therefore be marked Done without completing the milestone. Add explicit investigation/remediation tasks, or defer these deliverables and update the MVP scope accordingly.
AGENTS.md reference: AGENTS.md:L156-L157
Useful? React with 👍 / 👎.
| counts for `Cosmere.Tracker.TestKit`, count of `[CosmereTrackerAutoData]`/ | ||
| `[InlineCosmereTrackerAutoData]` call sites across the 18 test files, | ||
| current `dotnet test` run time, and a short readability note per |
There was a problem hiding this comment.
When the before/after runs differ in restore state, build-cache warmth, host load, or configuration, a single dotnet test runtime can produce an apparent regression or improvement unrelated to Compono. Because Phase 4 uses this evidence to decide whether performance regressed unacceptably, specify a fixed command/environment, separate build or restore time, collect repeated runs, and define how the comparison and acceptable threshold are calculated.
AGENTS.md reference: AGENTS.md:L32-L34
Useful? React with 👍 / 👎.
| **Implements:** [ADR-0029](../adr/0029-milestone-7-dogfooding-strategy-and-capability-gap-decision-framework.md) | ||
| (dogfooding strategy, migration-driven evidence, gap decision rubric, | ||
| where outcomes get recorded) |
There was a problem hiding this comment.
This roadmap plan lists only ADR-0029, but its scope explicitly executes ADR-0011, ADR-0018, ADR-0022, and ADR-0025 and may amend three of them. docs/plans/README.md requires a milestone plan's Implements line to list every accepted ADR it draws on, so the current header leaves the plan's dependency and decision trail incomplete; include the governing ADRs used by the migration.
Useful? React with 👍 / 👎.
Expands ADR-0029/PLAN-0007 per review: prefer idiomatic Compono over mechanical translation (documenting removed abstractions); treat migration friction as valid evidence even when a working Compono alternative exists; broaden evidence collection past line counts/runtime; require capturing positive findings for a balanced assessment; make the migration guide and an evidence-backed roadmap first-class living deliverables instead of after-the-fact write-ups; replace the binary roadmap/intentional-difference outcome with a five-way classification (bug / roadmap candidate / acceptable alternative / intentional design difference / migration-only friction) so blocking bugs get fixed through the normal engineering workflow instead of being designed around; and require Phase 4 to state a final architectural conclusion about whether dogfooding changed Compono's design direction. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
cosmere-tracker's AutoFixture kit has no semantic-data concept, so migration-driven evidence alone would never surface Compono.Bogus - the package would otherwise stay validated only by its own sample project. ADR-0029 now carves out Compono.Bogus as the one deliberate exception to "only migrate what's there": the migration must survey the domain models for realistic-data candidates and adopt UseBogus()/ADR-0028's alias/convention mechanism against cosmere-tracker's book/character/world vocabulary, which the built-in convention allowlist wasn't designed around - a genuine test of that mechanism, not just the happy path. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
1 participant
Summary
ncipollina/cosmere-tracker's AutoFixture test kit, and a four-question rubric (frequency, workaround cost, principle alignment, readability delta) for deciding each candidate capability gapcosmere-tracker's real fixture factory:Freeze<T>()-style hidden shared values, NSubstituteConfigureMembers, andOmitOnRecursionBehavior— none pre-approved as future features, each decided from evidence per ADR-0029Proposedroadmap ADR or a dated Amendment to the governing existing ADR (ADR-0011/0022/0025) recording "intentional design difference" — so a "no" is a real, discoverable outcome, not silencedocs/mvp.md's Milestone 7 section and the ADR/plan indexesTest plan
🤖 Generated with Claude Code