# @deksden-com/dd-flow-cli ## 0.8.0 ### Patch Changes - a092c98: Accept controlled OpenCode lifecycle events and cumulative root/child Session usage through the shared harness evidence path. - fb2395a: Accept trusted lifecycle events from controlled zcode-acp sessions and retain harness-native root, child, parent, and execution-daemon identities in Session evidence. - Promote the accepted SPECIFY-through-CODE-REVIEW runtime as stable 0.8.0 and align its shared compatibility, flow, stage-start and status schemas with Memory Bank 4.0.0. ## 0.8.0-beta.138 ### Patch Changes - Replace CODE and repair file allowlists with optional `planned_write_areas`: soft coordination hints that serialize likely concurrent collisions but never deny a necessary project-local edit. Record unexpected actual paths as `coordination.drift` instead of rejecting the Work. - Render the hard RUN workspace boundary separately from required reads, discovery hints and planned coordination areas, so a worker can distinguish semantic constraints from planning predictions. ## 0.8.0-beta.137 ### Patch Changes - Transfer a running coordinator Work to a fresh Session at a frozen `new_session` stage boundary, while still rejecting a second Session that tries to repeat the same stage. ## 0.8.0-beta.136 ### Patch Changes - Serialize the long CODE aggregate gate and reuse complete same-workspace receipts after a client loses its command output, preventing a concurrent retry from duplicating checks or colliding on receipt identifiers. ## 0.8.0-beta.135 ### Patch Changes - Derive a CODE repair Work's additional write scope from exact existing project files named by its retained failed-check receipt, with path-escape and dependency-directory guards. - Keep eval snapshot rebasing out of restored workspaces and installed tools, so valid JSON-with-comments project files cannot break a later stage entry. ## 0.8.0-beta.134 ### Patch Changes - Pair the engine with the PLAN/PLAN-REVIEW prompt correction that makes every linked acceptance scenario explicitly owned for update or explicitly classified as unaffected before CODE opens. ## 0.8.0-beta.133 ### Patch Changes - Reject a PLAN that assigns a project aggregate gate to one scoped CODE Work; aggregate gates now run at the CODE/readiness fan-in, where failure can create a correctly scoped repair Work. - Preserve the distinction between a worker that explicitly failed its Work and a worker that omitted its lifecycle finish in eval runner evidence. ## 0.8.0-beta.132 ### Patch Changes - Preserve the one-to-one relationship between PLAN-declared CODE checks and their execution receipts, even when checks intentionally share a command. - Reject a CODE-REVIEW decision that rejects or defers a material P0/P1 finding before repair Work is created, so the initial repair packet is complete and cannot leave the stage in a non-progressing state. ## 0.8.0-beta.131 ### Patch Changes - Explain that a CODE Work receives shared end-to-end acceptance context but owns only its declared contribution, write scope and checks. This prevents a correctly scoped worker from failing merely because a dependent UI or documentation surface belongs to another Work. ## 0.8.0-beta.128 ### Patch Changes - Render an executable, one-command heredoc template for every agent-owned HITL pause and preserve that template through Codex, ZCode, Grok, OpenCode and Antigravity hook parsing. ## 0.8.0-beta.127 ### Patch Changes - Keep PROTOCOLIZE stage attachment independent from coordinator Session reuse, expose the measured worker capacity in the PLAN packet, and report an explicit zero-capacity infrastructure outcome for PLAN-REVIEW. ## 0.8.0-beta.122 ### Patch Changes - Preserve a portable Git bundle in each new eval snapshot and restore it before the checkpoint tree, so later stages retain their branch and can create managed worktrees. ## 0.8.0-beta.121 ### Patch Changes - Keep CODE implementation in fresh Work sessions so the coordinator preserves graph, evidence and final-gate context across a sequential delivery graph. ## 0.8.0-beta.119 ### Patch Changes - Keep the accepted `@check/...` declaration when creating a CODE repair Work from a failed receipt. Receipts retain the expanded command as evidence, but the repair packet is validated and executed through the original alias. ## 0.8.0-beta.118 ### Patch Changes - Keep PLAN-time document baselines immutable while validating CODE handoff, so a required durable-documentation Work cannot invalidate its own generated CODE batch. - Emit full Work ids in generated lifecycle commands and align the CODE verification example with the installed `code-verification@2` schema. ## 0.8.0-beta.116 ### Patch Changes - Register the CODE stage before workspace readiness so a failed bootstrap is a recoverable, timed stage attempt rather than an unfinishable lifecycle. ## 0.8.0-beta.115 ### Patch Changes - Materialize both partially filled PLAN artifacts at stage start and report which fields are already owned by the CLI. - Prepopulate the aspect-map catalog identifiers without making semantic applicability decisions for the planner. - Return exact artifact paths and schema-validation commands in the PLAN packet while preserving `plan.json` as the sole semantic plan. ## 0.8.0-beta.111 ### Patch Changes - Make a future CODE check an explicit, provider-owned `@check/...` alias with an exact definition; reject unordered, raw or unmaterialized checks. - Run PLAN-declared available readiness checks after bootstrap, preserve their receipts, and distinguish project policy gates from planner-selected CODE checks. - Carry the stricter verification contract through PLAN and CODE worker packets without asking the CLI to judge test relevance or cost. ## 0.8.0-beta.110 ### Patch Changes - Make PLAN own verification selection through a single check catalogue; validate only references, provider ordering and guarded commands. - Preserve review evidence in Work results instead of mutating aspect maps, reuse one RUN capacity measurement, and support HITL-backed effective obligation amendments. - Never rewrite Bash heredocs in lifecycle hooks. ## 0.8.0-beta.109 ### Patch Changes - Keep the SQLite vNext RUN projection free of legacy `current_stage`, attempt, worker and coverage mirrors; derive stage guidance from canonical stage runs. ## 0.8.0-beta.108 ### Patch Changes - Align vNext runtime identity, Work/Session state, filesystem materialization, stage reports, review findings and check receipts with SPC-009. - Add a single stage catalog, safe RUN path resolver, overlapping write-scope scheduling guard and structured eval snapshot rebasing. - Remove automatic Work `try-001` materialization and the separate `work.json` projection; explicit retries archive under `attempts/ATT-NNN`. ## 0.8.0-beta.103 ### Patch Changes - Isolate local CODE check processes by checkout and pass the scope through deterministic bootstrap and gate commands. - Make CODE worker instructions distinguish repairable project check failures from evidenced external blockers. ## 0.8.0-beta.102 ### Patch Changes - Permit the stdin pipe required by `stage resume` while retaining the strict standalone-command guard for stage and Work starts. ## 0.8.0-beta.101 ### Patch Changes - abef4b2: Reject ambiguous compound lifecycle starts, preserve trusted coordinator ownership, enforce hard Work dependencies during review dispatch, and expose complete reviewer, timing, evidence, Session and Agent observability. ## 0.8.0-beta.100 ### Patch Changes - 430e49f: Preserve project check aliases until execution, add recoverable external stage blockers, deduplicate physical-session usage, and improve session and review routing projections. - 06e58ab: Stream machine-readable progress for long lifecycle commands, make CODE and CODE-REVIEW retries idempotent, preserve terminal timestamps, and tighten candidate usage/session projections. - Reject a completed CODE verification that cites evidence missing from the RUN or its workspace, and refresh protocol summaries at deterministic CODE boundaries. ## 0.8.0-beta.99 ### Patch Changes - Validate draft 2020-12 stage contracts with the matching JSON Schema engine and surface schema-registry compilation failures as actionable engine errors. - Namespace CODE-REVIEW findings by reviewer group and require every assigned aspect exactly once before accepting reviewer evidence. - Keep active RUN projections free of resolved pause/blocker records, preserve their history in the timeline, and normalize quoted RUN ids from hooks. - Validate CODE completion evidence before running the aggregate project gate. ## 0.8.0-beta.93 ### Patch Changes - Republish the beta.92 CODE packet validation fix as the matched engine for the corrected flow-pack compatibility declaration. ## 0.8.0-beta.92 ### Patch Changes - Treat each PLAN `verification.checks` entry as an executable command only. With a project check profile, PLAN now accepts only declared aliases or exact declared commands, and rejects semantic assertions before it can create a CODE Work packet. Semantic proof stays in `expected_evidence`. ## 0.8.0-beta.91 ### Patch Changes - Remove obsolete `session_coverage` and `usage_coverage` projections from vNext RUN state, leaving factual Session and usage data to the stat commands. ## 0.8.0-beta.90 ### Patch Changes - Snapshot one project-owned execution profile into each vNext RUN and keep checkout routing separate from handoff, review and bootstrap policy. - Move deterministic workspace bootstrap to CODE, report its heartbeat, and end successful CODE at the truthful terminal verdict `code_completed`. - Use readable `WRK-*` Work identifiers and replace synthetic coverage claims in vNext stage reports with factual post-settlement statistics commands. ## 0.8.0-beta.89 ### Patch Changes - Keep active CODE workers alive until the harness reports a terminal state. - Stream long check progress and validate project-owned check aliases before CODE Work registration. - Scope Session, token, tool-call and timing evidence to the selected stage or segment and keep only the latest check receipt as current evidence. ## 0.8.0-beta.88 ### Patch Changes - Preserve accepted requirements, semantic responsibility, selected project context, write scope and verification when PLAN projects CODE Work. - Execute CODE as the registered Work graph under the RUN root orchestrator, with focused checks at Work completion and a deterministic aggregate gate at stage completion. - Keep failed aggregate gates open and create an explicit repair Work from the failed receipt plus the original Work context. ## 0.8.0-beta.87 ### Patch Changes - Remove the obsolete vNext fixture handoff fallback so PROTOCOLIZE starts only from the accepted SPECIFY context. - Prove that review-off uses the same validated final PLAN-to-CODE closure as an enabled review. ## 0.8.0-beta.86 ### Patch Changes - Make SPECIFY JSON the accepted semantic source, preserve all R/AC obligations through PROTOCOLIZE and PLAN, and derive the CODE work batch only after shared deterministic validation. - Reject stale PLAN/CODE handoffs and publish accepted semantic artifacts and deterministic reports atomically. - Resolve Codex transcripts inside the selected `CODEX_HOME`; tests now use isolated homes instead of scanning a developer's session history. ## 0.8.0-beta.81 ### Patch Changes - Reject an accepted PLAN-REVIEW correction unless it regenerates the proposed CODE batch and records that artifact in the correction receipt. This keeps CODE work paths and verification aligned with the revised plan. ## 0.8.0-beta.80 ### Patch Changes - Make failed eval snapshots read-only: an accidental worktree value for `--project-root` no longer registers a second project in the dedicated runtime before failing. ## 0.8.0-beta.79 ### Patch Changes - Use the single vNext workspace policy for the frozen SPECIFY → PROTOCOLIZE handoff. This prevents the generic runtime default from contradicting `project-workspace.json`. - Remove the obsolete `execution.stage_handoff` runtime setting so it cannot be mistaken for a second vNext authority. ## 0.8.0-beta.78 ### Patch Changes - Treat a complete PLAN-REVIEW reviewer result with `blocked` as actionable evidence, requiring a classified in-place correction, rather than a terminal stage failure. Missing, malformed, or unfinished reviewer evidence still blocks the stage. - Keep the PROTOCOLIZE PSET test fixture consistent with the fail-closed acceptance-coverage contract. ## 0.8.0-beta.77 ### Patch Changes - Make PROTOCOLIZE fail closed unless `acceptance_coverage` owns every `AC-*` criterion in the accepted SPECIFY handoff exactly once. ## 0.8.0-beta.76 ### Patch Changes - Make the PROTOCOLIZE workspace receipt a fail-closed gate for PLAN, PLAN-REVIEW and CODE, and include the verified write workspace in every generated worker packet. - Require disposable capacity probes and completed review workers to release their harness slots before productive work continues. ## 0.8.0-beta.75 ### Patch Changes - Render every vNext lifecycle command with the bound `DD_FLOW_HOME`, so a worker can finish or hand off a stage from any shell environment. ## 0.8.0-beta.74 ### Patch Changes - Measure fresh-subagent capacity at PLAN-REVIEW dispatch rather than reusing a stale PLAN observation; PLAN now groups review aspects semantically only. ## 0.8.0-beta.73 ### Patch Changes - Require self-contained CODE work batches at PLAN finish. ## 0.8.0-beta.72 ### Patch Changes - Align the validated flow contract with the vNext route: a run-scoped feature worktree is provisioned at PROTOCOLIZE start, before durable protocol files are materialized. ## 0.8.0-beta.71 ### Patch Changes - Keep the worktree route strict while allowing Desktop harnesses whose task cwd cannot be dynamically changed: stage packets use absolute workspace targets and hook cwd remains audit evidence rather than a false blocker. ## 0.8.0-beta.70 ### Patch Changes - Provision the configured feature worktree and its bootstrap deterministically at vNext PROTOCOLIZE start. PROTOCOLIZE finish now only validates the frozen route and materializes durable documents there; PLAN and later writing stages reject a session started outside the recorded workspace. ## 0.8.0-beta.58 ### Patch Changes - Bind a portable fixture's synthetic root Work to the first real stage session before starting its child PLAN Work, so a focused PLAN run retains the normal parent-session invariant without importing old sessions. ## 0.8.0-beta.57 ### Patch Changes - Preserve the final newline of `git diff --binary` in exported portable fixtures so their project overlays can be imported by `git apply`. ## 0.8.0-beta.56 ### Patch Changes - Report PROTOCOLIZE as the next legal stage after finished SPECIFY; PLAN is now guided only after the PROTOCOLIZE stage is complete. ## 0.8.0-beta.55 ### Patch Changes - Keep a finished SPECIFY stage isolated until its returned PROTOCOLIZE start command is explicitly invoked; completion now records the hash of the actual stored document and exposes acceptance/check checks in the stage report. ## 0.8.0-beta.54 ### Patch Changes - Restore the accepted fixture's PLAN-REVIEW setting alongside its runtime variables so isolated targets retain the configured flow behavior. ## 0.8.0-beta.53 ### Patch Changes - Mark engine-exported stage fixtures as accepted, matching the strict controller import gate. ## 0.8.0-beta.52 ### Patch Changes - Add an operator-only portable stage-fixture export/import boundary for stage-isolated evaluations. It restores accepted predecessor artifacts and RUN variables into a fresh runtime, never old Sessions, Works, usage or machine paths. ## 0.8.0-beta.51 ### Patch Changes - Put the measured per-wave fresh-subagent capacity into the PLAN response, rendered prompt and Work context. PLAN now explicitly minimizes review waves before reviewer count while preserving real semantic boundaries. ## 0.8.0-beta.49 ### Patch Changes - Prefix generated vNext inter-stage start commands with the selected `DD_FLOW_HOME`, so their PreToolUse hook binds the same isolated RUN. - Allow a bounded capacity observation to replace an earlier observation when the harness completes its probe window with a different factual count. ## 0.8.0-beta.48 ### Patch Changes - Publish the coordinated engine snapshot for the vNext HITL flow-pack beta.48. ## 0.8.0-beta.47 ### Patch Changes - Make user interaction an in-stage pause/resume lifecycle: preserve the same Work, stage and attempt, store raw question/answer intake, and require a trusted resume hook event. - Remove vNext `waiting_for_user` finish outcomes and fictitious PROTOCOLIZE-to-SPECIFY answer routes. Generated prompts now return exact pause/resume commands, including an isolated `DD_FLOW_HOME` where needed. ## 0.8.0-beta.45 ### Patch Changes - Capture the immutable PLAN-REVIEW decision receipt only after a finish payload passes the mechanical gate, so rejected attempts remain retryable. ## 0.8.0-beta.44 ### Patch Changes - Record RUN capacity before PLAN and expose one consistent `run capacity record` command with the bounded 15-probe contract. - Preserve the proposed CODE batch through PLAN-REVIEW, validate it before PLAN acceptance, and keep the temporary reviewer dispatch file ephemeral. - Keep per-Session stage state accurate, preserve immutable PLAN-REVIEW decision payloads, and materialize reviewer findings into the aspect map. - Render a useful deterministic PLAN-REVIEW Markdown/HTML report and require stable finding ids for mechanical decision coverage. ## 0.8.0-beta.43 ### Patch Changes - Attribute transcript usage by the linked Work/Session id; Codex child transcripts expose the child id in `session_meta.id` while the provider id can remain the parent thread id. - Include initial/final PLAN revisions and checksums, correction paths and `semantic_re_review` in the deterministic PLAN-REVIEW report. ## 0.8.0-beta.42 ### Patch Changes - Make PLAN-REVIEW a single semantic review wave: reviewer `needs_changes` is evidence, while the orchestrator applies accepted fixes in the same Work and finishes once with an explicit correction receipt. - Remove automatic review retries and validate only mechanical correction coherence; semantic re-review remains `not_run_by_policy`. ## 0.8.0-beta.41 ### Patch Changes - Make capacity probing an explicit leaf-session measurement: probes wait 60 seconds, return `AGENT-NN`, use a 180-second total deadline, and count only exact-token completions. ## 0.8.0-beta.40 ### Patch Changes - Include the isolated `DD_FLOW_HOME` prefix in generated worker completion and failure commands, matching the worker-start command in a fresh Desktop session. ## 0.8.0-beta.39 ### Patch Changes - Replace partial PLAN-review retries with a new complete review attempt after any PLAN revision. The completed attempt is archived by the common stage state machine; stale evidence cannot be reused. - Return the isolated runtime prefix in reviewer worker-start commands, so a fresh Desktop worker reaches the same RUN without manual environment setup. ## 0.8.0-beta.38 ### Patch Changes - Treat fresh-session capacity as one bounded harness observation rather than fifteen registered probe Works. `plan-review capacity record` stores the observed slot count once for the RUN and lets reviewer dispatch proceed. ## 0.8.0-beta.33 ### Patch Changes - Replace beta-only Agent Turn, launch-token and adapter-binding mechanics with hook-observed Work/Session links; add RUN variables, fresh-review capacity probes and transcript-source usage recomputation. ## 0.8.0-beta.32 ### Patch Changes - Split independent PLAN review into a first-class `plan-review` stage with an atomic CODE opening gate and RUN-scoped review policy. ## 0.8.0-beta.31 ### Patch Changes - Let PLAN dispatch a new review attempt only for groups whose latest completed verdict is `needs_changes`; PLAN acceptance uses the latest attempt for each group. ## 0.8.0-beta.30 ### Patch Changes - Include the explicit project root in every generated worker start command, so a Desktop worker launched outside the target checkout can be observed and claimed by its Codex hook. ## 0.8.0-beta.29 ### Patch Changes - Give each Work a one-time launch token, claim the exact matching PreToolUse event, and let the trusted Desktop harness bind a returned child-task identity before a parallel worker starts. ## 0.8.0-beta.28 ### Patch Changes - Keep vNext RUNs live across same-session stage handoffs, dispatch grouped PLAN reviews as real Work/Agent Turns before PLAN acceptance, and preserve portable RUN references in semantic artifacts and worker packets. ## 0.8.0-beta.27 ### Patch Changes - Materialize grouped PLAN aspect routes as prerequisite Work, normalize RUN source provenance, and report PLAN guidance and artifacts without external filesystem-relative paths. ## 0.8.0-beta.26 ### Patch Changes - Keep PreToolUse hook rewriting scoped to one `dd-flow` invocation; compound shell commands are recorded but never mutated. ## 0.8.0-beta.25 ### Patch Changes - Route a Desktop PreToolUse event to an explicitly prefixed isolated `DD_FLOW_HOME`, so beta evals keep trusted session binding without touching the shared runtime database. ## 0.8.0-beta.24 ### Patch Changes - Unify vNext Work/Turn authority, bind PLAN/CODE to trusted hook sessions, enforce strict SPECIFY-to-PLAN traceability, and emit portable Work/usage evidence with deterministic stage reports. ## 0.8.0-beta.13 ### Patch Changes - Automatically apply safe, additive internal SQLite schema updates whenever a write-capable engine command opens an existing runtime database. This keeps hook-event binding compatible across beta engine upgrades without treating it as a Memory Bank data migration. ## 0.8.0-beta.12 ### Patch Changes - Bind vNext SPECIFY bootstrap to a one-time, fresh PreToolUse event using a canonical invocation fingerprint when Codex Desktop ignores hook `updatedInput` rewrites. ## 0.7.0 ### Minor Changes - Add target-driven engine routing, explicit storage migration safety, immutable RUN engine bindings, engine schema registry validation, and self-contained plan execution guidance. ## 0.6.0 ### Minor Changes - 47cd270: Add SPC-006 bootstrap stage start/context packets and sealed stage finish, while repairing Memory Bank upgrade stage routing, Codex hook event identity, PATH-stable per-home hook installation, canonical Codex session registration, session coverage projections, terminal orchestrator closure, and safe reconciliation of runs whose legacy authority and workspace are gone. ## 0.5.0 ### Minor Changes - 43bf9c2: Cut over runtime execution to the canonical SQLite-backed flow contract with a single PreToolUse binding, canonical plan progress, resumable worker coverage, and validated `stage-report.json`, `stage-report.md`, and `stage-report.html` artifacts. ## 0.4.2 ### Patch Changes - Preserve explicit `--project-root` and `--root` options for top-level router commands, so engine selection uses project compatibility instead of an unrelated globally installed snapshot. ## 0.4.1 > Release-accounting correction: the entries below were already present in > the published `0.4.1` tarball built from `6f85c04`, but their pending > Changesets had not been consumed. ### Minor Changes - Add `dd-flow prompt render` for deterministic RUN-local worker prompts, including validated semantic context and checked `run://` reads. ### Patch Changes - Preserve `--project-root` through engine routing, compatibility preflight, protocol services, nested lookups, merge-queue commands, and post-mutation dashboard refresh. Duplicate rootless protocol ids fail closed without side effects, while successful mutations remain successful when an optional dashboard projection fails. - Include production dependencies in installed engine snapshots, so a routed CLI works outside the source checkout. - Expand `dd-flow run timeline` with stage durations, session roster, usage coverage, artifact references and repeatable `--hide` controls. - Update canonical Memory Bank resolution for the `mb-upgrade` `05-review` stage instead of requiring a separate top-level `mb-upgrade-review.md`. - Make typed `PRT-*` and `RUN-*` identity project-scoped and monotonic across registered projects, migrate legacy global keys without renumbering, and ignore legacy date-based protocol ids during allocation. - Add source-aware flow-flag snapshots, provenance revisions, grouped session coverage and privacy-safe timing/usage projections. - Validate the complete flow-contract catalog before mutable RUN/protocol starts and expose exact read-only diagnostics for invalid catalogs. ## 0.4.0 ### Minor Changes - Add branch/workspace protocol context for agents: `protocol branch-status`, `branch_context` in protocol and merge JSON outputs, and branch bundle status with eligible, not-ready and claimed protocol lists. Add `merge bundle status|claim|complete|fail` commands so a merge worker can integrate one feature branch containing multiple ready protocols under the existing merge lane lock. ### Patch Changes - cfdcf1d: Fix compatibility status reporting so selected compatible engines advertise `normal_write` in `allowed_modes`; blocked operations are now reported only when a write is actually blocked by missing or incompatible engine resolution. ## 0.3.1 ### Patch Changes - 16f68a3: Add router-native engine snapshot commands and routed project-command dispatch. Also include the runtime/dashboard compatibility wave needed by the current Memory Bank canon: - target-based dashboard `open`/`refresh` commands and related-command help; - persisted FIFO lane waiters and merge wait-acquire specialization; - `migration plan/report/verify` evidence helpers for controlled `mb-upgrade` runtime migrations; - `dd-flow/project-summary@1` publication and global dashboard aggregation from project summaries. - active npm release runbook notes for Changesets-generated changelogs, strict Memory Bank canon build metadata, token-safe publish, local global link and post-publish compatibility verification. ## 0.3.0 ### Minor Changes - Add CLI version preflight support with `dd-flow version`, `dd-flow --version`, `dd-flow version --json`, Memory Bank compatibility verdicts in `dd-flow status`, and an explicit cached npm registry check through `dd-flow status --check-registry`. - 565c601: Add static HTML dashboard rendering for global, project, and protocol dashboard pages. The dashboard command family now supports `--format html`, a `dashboard data` command, adjacent JSON data files, protocol pages, bundled dashboard schemas, and compatibility with existing markdown dashboards. - Add `flow_guidance` to status-like protocol/run/project/merge outputs so agents can see current stage, allowed next stages, recommended prompt/action, guard status and missing predecessor evidence without moving semantic planning into the CLI. - Store new `RUN-*` artifacts in DD Flow home under `~/.dd-flow/projects//runs//`, while keeping legacy project `.tasks/dd-flow-runs//` readable as compatibility input. - Add protocol-set continuation support: protocol dependency blockers, `protocol ready`, safe `protocol-implement` guidance, blocked-protocol start guards and compact set status data for dashboards. - Add `mb-sdlc-review` / `review` flow-kind support, bundled `mb-sdlc-review-report` schema validation, semantic review-report checks and dashboard counters for review runs. - Extend dashboard data/rendering for protocol sets, review runs and home-run stage artifacts. ### Patch Changes - Keep merge queue, protocol runtime and run status in sync when a merge job completes, so agents can verify `closed` protocol state, `merged` queue state and `done/merged` run state without stale `claimed` diagnostics. ## 0.2.0 ### Minor Changes - Expose canonical Memory Bank version visibility in `dd-flow status`, including CLI build metadata, resolved canon metadata, project Memory Bank markers, project flow pack manifest state, drift diagnostics, and the `dd-flow/status-report@1` schema. ### Patch Changes - 322cae9: Add canonical Memory Bank discovery commands and runtime status output.