You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Story-scoped workflows take a v7 ticket directly. Automate red and expand mode, Trace, NFR and Test Review accept a story the way the method names it (1.2, a title, a leaf file), read it through the method's tickets.py find without needing the entry pulled, and fall back to the v6 story file only for v6 input. One shared contract, bmod-tea/contracts/story-identity.md, names every story output by its ticket (shop--e1-s2), so a story's checklist, summary, NFR report, review and matrix share one key that never collides across initiatives and never changes on a retitle. Red mode halts on criteria it cannot turn into failing tests, hands the builder the story's plan path, and links artifacts back only into a v6 story file, never a ticket.
bmad migrate tea carries story outputs written before v7 keys onto their tickets. After bmad migrate method, it maps each checklist, summary, matrix and report to its ticket from exact evidence (its story file, the entry's v6_key, or the method's migration record), shows a plan, and on approval renames each file and rewrites its identity fields. Historical citations stay as written beside a resolved-source list, nothing is overwritten, conflicts wait for the user, and anything without exact evidence stays in place as unresolved. It also covers native v7 projects whose outputs were keyed by a leaf file name.
The bmad skill can now route to TEA inside the BMad v7 method flow. TEA's module help says where each workflow fits around bmad-build: epic Test Design and red mode before a story is built, expand mode after it for a maintained suite (with bmad-qa-generate-e2e-tests still the fit for a quick one-off), and Test Review, NFR and Trace before a release. Method skills are named only when installed, so the routes read correctly with TEA alone. The help also tells bmad how to see what TEA has run: it resolves modules.tea.test_artifacts, lists the per-workflow folders and the older root-level file names, and treats an in-progress checkpoint as a started run.
Fixed
Test Design reads the v7 ticket tree where bmad migrate leaves it, in the initiative's tickets/ folder. An initiative still in the preview layout, with tickets.toml, epic folders, leaf files or plans directly inside it, now stops automatic mode selection and points the user at bmad migrate; an explicit system- or epic-level request still proceeds.
The red-mode ATDD checklist no longer tells users to mark a story done in its tickets.toml entry, which holds no status. It now covers a v7 ticket (the builder records the work in the story's plan, and the user or orchestrator has the Product Owner mark it done, through tickets.py mark on a repo store or a transition on a tracker store), a v6 story, and projects without the BMad method module, and states that red mode changes no status.
Evaluate hands its CI plan to bmad-tea-framework without naming files inside that skill, so it no longer breaks if Framework's internals move.
The write-time enforcement hook that bmad-tea-framework installs no longer blocks two kinds of valid write, and catches two violations it missed. C3 reads assertion operands with their string literals intact, so expect(result['actual']).toEqual(result['expected']) and other operands that differ only inside a string are no longer reported as tautologies. The pre-write check applies an Edit or MultiEdit to the file on disk and scans the result, so editing text inside an existing comment is no longer blocked, and it blocks only on violations the edit introduces. The file-wide rules, the line limit and the Maestro no-assertion check, now also run on edits. C2 and C3 match across line breaks, so a formatter-split test.only\n(...) or expect(actual)\n .toBe(actual) is now caught. Projects that already installed the hook pick up the fix by re-copying .claude/hooks/tea-enforce.cjs and updating hookSha256 in .tea/enforce-config.json (#449).
Changed
The Evaluate how-to guides state each guide's purpose up front before explaining mechanics. The three first-run guides (skill or agent, stdio MCP server, and HTTP API) open with explicit bottom lines and the core workflow sequence (confirming requirements, authoring, validating with check, qualifying mutations with preflight, executing trials with run, and interpreting evidence with score). Task-specific guides open with their focused responsibilities and retain their narrower approval boundaries, while the repeatability guide preserves its offline-first saved-records scope.
The Run Test Review in CI guide opens its GitHub Actions section with the tea-test-review GitHub Action, the one-step route, before the self-managed workflow template.
The CLI reference pages name every supported agent the same way. Each one lists Claude Code, Codex, Antigravity and --agent custom up front and says the run uses the session that agent CLI already holds, so there is no login step before a run. The Test Design CLI quickstart no longer tells readers to run codex login, and the Test Review prerequisites describe signing in to an agent as one-time setup done with its vendor.