Skip to content

Alignment Review: repo

Alignment Review: repo

Review Metadata

Review Date: 2026-03-28 Scope: repo Status: complete Review Epic: hx-1297bf62 Review Issues: hx-a3dab13b, hx-5c8aee1e, hx-c41039f2, hx-3eb44502 Primary Governing Artifact: workflows/README.md

Scope and Governing Artifacts

Scope

  • planning stack and workflow contract
  • helix wrapper CLI and run-loop behavior
  • built-in tracker implementation and queue hygiene
  • HELIX skill packaging, installer, and validation surfaces

Governing Artifacts

  • docs/helix/00-discover/product-vision.md
  • docs/helix/01-frame/prd.md
  • docs/helix/01-frame/features/FEAT-001-helix-supervisory-control.md
  • docs/helix/01-frame/features/FEAT-002-helix-cli.md
  • docs/helix/02-design/adr/ADR-001-supervisory-control-model.md
  • docs/helix/02-design/adr/ADR-002-tracker-write-safety-model.md
  • docs/helix/02-design/contracts/API-001-helix-tracker-mutation.md
  • docs/helix/02-design/solution-designs/SD-001-helix-supervisory-control.md
  • docs/helix/02-design/technical-designs/TD-002-helix-cli.md
  • docs/helix/03-test/test-plans/TP-002-helix-cli.md
  • docs/helix/04-build/implementation-plan.md
  • workflows/README.md
  • workflows/EXECUTION.md
  • workflows/TRACKER.md
  • workflows/actions/reconcile-alignment.md
  • workflows/templates/alignment-review.md

Intent Summary

  • Vision: HELIX should act as a supervised autopilot that advances the weakest ready layer while preserving authority order, bounded execution, and tracker-first discipline (docs/helix/00-discover/product-vision.md:5-11, docs/helix/00-discover/product-vision.md:36-40, docs/helix/00-discover/product-vision.md:44-69).
  • Requirements: helix run should choose the least-power next bounded action, react to downstream artifact changes, preserve tracker-backed execution, and keep the packaged skill surface coherent (docs/helix/01-frame/prd.md:14-18, docs/helix/01-frame/prd.md:82-130, docs/helix/01-frame/prd.md:151-183).
  • Features / Stories: FEAT-001 defines supervisory control and direct intervention points; FEAT-002 defines the CLI surface, tracker model, installer contract, and deterministic wrapper proof lane (docs/helix/01-frame/features/FEAT-001-helix-supervisory-control.md:35-47, docs/helix/01-frame/features/FEAT-001-helix-supervisory-control.md:60-90, docs/helix/01-frame/features/FEAT-002-helix-cli.md:23-115).
  • Architecture / ADRs: ADR-001 selects the supervisory control model; ADR-002 selects conservative file-backed tracker safety and visible concurrency semantics (docs/helix/02-design/adr/ADR-001-supervisory-control-model.md:23-39, docs/helix/02-design/adr/ADR-002-tracker-write-safety-model.md:24-40).
  • Technical Design: SD-001 and TD-002 require bounded queue-drain routing, explicit PLAN and POLISH handling, queue-drift revalidation, review-aware stopping, and coherent package-relative skill installation (docs/helix/02-design/solution-designs/SD-001-helix-supervisory-control.md:16-31, docs/helix/02-design/solution-designs/SD-001-helix-supervisory-control.md:67-103, docs/helix/02-design/technical-designs/TD-002-helix-cli.md:28-40, docs/helix/02-design/technical-designs/TD-002-helix-cli.md:75-101, docs/helix/02-design/technical-designs/TD-002-helix-cli.md:182-247).
  • Test Plans: TP-002 defines bash tests/helix-cli.sh as the deterministic verification path for wrapper and tracker behavior (docs/helix/03-test/test-plans/TP-002-helix-cli.md:8-30, docs/helix/03-test/test-plans/TP-002-helix-cli.md:40-72).
  • Implementation Plans: the 04-build layer is supposed to sequence the current tracker-backed slices, but the current plan still cites superseded issues instead of the live follow-up set (docs/helix/04-build/implementation-plan.md:13-16, docs/helix/04-build/implementation-plan.md:46-52, docs/helix/04-build/implementation-plan.md:70-74).

Planning Stack Findings

FindingTypeEvidenceImpactReview Issue
FEAT-002 still documents the older six-code queue-drain contract and omits PLAN and POLISH, while the maintained workflow and technical design treat them as first-class NEXT_ACTION outcomes.staledocs/helix/01-frame/features/FEAT-002-helix-cli.md:44-57, docs/helix/01-frame/features/FEAT-002-helix-cli.md:102-109, docs/helix/02-design/technical-designs/TD-002-helix-cli.md:75-90, workflows/EXECUTION.md:65-82High: the governing feature spec no longer traces cleanly to the implemented execution contract.hx-a3dab13b
TP-002 still cites generic workflow references for tracker metadata coverage instead of the specific tracker contract and harness sections that actually prove those behaviors.staledocs/helix/03-test/test-plans/TP-002-helix-cli.md:96-106, workflows/TRACKER.md:90-139, tests/helix-cli.sh:1945-2014Low: the test plan remains usable, but its evidence chain is weak.hx-a3dab13b
docs/helix/04-build/implementation-plan.md still sequences superseded issue IDs hx-f8bfa352, hx-c2c1557c, and hx-a039f874 instead of the live follow-up set.staledocs/helix/04-build/implementation-plan.md:46-52, docs/helix/04-build/implementation-plan.md:70-74, ddx bead list --status open --json on 2026-03-28Medium: the 04-build layer exists, but it has drifted from the actual queue and no longer provides current sequencing authority.hx-a3dab13b

Implementation Map

  • Topology: workflows/ holds the normative workflow contract and shared resources; scripts/helix is the wrapper CLI; scripts/tracker.sh is the file-backed tracker; .agents/skills/ is the project-level package surface; skills/ holds the canonical skill directories; tests/helix-cli.sh and tests/validate-skills.sh are the deterministic proof surfaces (workflows/README.md:17-40, scripts/helix:15-22, scripts/tracker.sh:7-18, tests/validate-skills.sh:4-18).
  • Entry Points: scripts/helix exposes run, implement, check, align, backfill, plan, polish, next, review, experiment, and tracker; tracker subcommands dispatch through tracker_dispatch; local installation writes ~/.local/bin/helix and links canonical project skills into ~/.agents/skills and ~/.claude/skills (scripts/helix:40-93, scripts/helix:522-643, scripts/helix:730-880, scripts/tracker.sh:761-813, scripts/install-local-skills.sh:36-45, scripts/install-local-skills.sh:92-110).
  • Test Surfaces: this pass reran bash tests/helix-cli.sh, bash tests/validate-skills.sh, and git diff --check; the wrapper harness passed all 77 deterministic cases and the skill validator reported 10 validated HELIX skills. The harness still lacks a queue-drain check-failure regression case for the run_check guard in scripts/helix, which leaves hx-7d00b06f as an open runtime test gap (docs/helix/03-test/test-plans/TP-002-helix-cli.md:12-16, tests/helix-cli.sh:1662-1707, tests/helix-cli.sh:2330-2350, scripts/helix:818-821, tests/validate-skills.sh:42-91).
  • Unplanned Areas: none found in the reviewed repo surface. Current gaps are stale planning layers, incomplete review follow-through, and missing regression coverage rather than orphaned implementation paths. This is an inference from the absence of uncovered features after revalidating the planning stack, runtime, tracker, installer, and proof lane in this pass.

Acceptance Criteria Status

Story / FeatureCriterionTest ReferenceStatusEvidence
FEAT-001 / US-001Given a repository with vision and PRD, when HELIX can safely continue, then helix run advances the next bounded layer without asking for a phase name.tests/helix-cli.sh:573-633SATISFIEDdocs/helix/01-frame/features/FEAT-001-helix-supervisory-control.md:65-71, scripts/helix:741-878, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-001 / US-001Given a user-requested functionality change, when it affects downstream artifacts, then HELIX routes to alignment or planning before implementation resumes.tests/helix-cli.sh:693-769SATISFIEDdocs/helix/01-frame/prd.md:151-165, docs/helix/02-design/solution-designs/SD-001-helix-supervisory-control.md:20-25, scripts/helix:832-859, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-001 / US-002Given a user invoking a specific layer directly, when they do so, then HELIX performs that action without breaking the supervisory model.tests/helix-cli.sh:531-550, tests/helix-cli.sh:1071-1114SATISFIEDdocs/helix/01-frame/features/FEAT-001-helix-supervisory-control.md:79-81, scripts/helix:522-643, scripts/helix:886-920, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-002Running helix help shows the command surface and key options.tests/helix-cli.sh:508-529SATISFIEDdocs/helix/01-frame/features/FEAT-002-helix-cli.md:27-40, scripts/helix:40-93.
FEAT-002Running ddx bead subcommands supports create/show/update/close/list, ready/blocked queries, dependency management, and status summaries.tests/helix-cli.sh:1758-2058SATISFIEDworkflows/TRACKER.md:90-139, scripts/tracker.sh:147-399, scripts/tracker.sh:402-594. Inference: the open create --help defect is a help-path safety bug, not evidence that the core tracker surface is absent.
FEAT-002Running helix run follows the explicit NEXT_ACTION contract.tests/helix-cli.sh:531-539, tests/helix-cli.sh:693-837SATISFIEDdocs/helix/02-design/technical-designs/TD-002-helix-cli.md:75-101, scripts/helix:818-878. The implementation supports PLAN and POLISH; the feature spec remains stale.
FEAT-002Running helix run does not attempt implementation after WAIT.tests/helix-cli.sh:2346SATISFIEDscripts/helix:861-863, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-002Running helix run stops and surfaces the exact backfill command after BACKFILL.tests/helix-cli.sh:2347SATISFIEDscripts/helix:865-868, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-002Running helix run counts only completed implementation passes as completed cycles.tests/helix-cli.sh:2348-2349SATISFIEDscripts/helix:730-813, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-002Running helix run surfaces review findings before the loop advances.tests/helix-cli.sh:1662-1707SATISFIEDdocs/helix/01-frame/features/FEAT-002-helix-cli.md:60-63, scripts/helix:772-789. Inference: this criterion is satisfied at the feature-spec layer because the loop stops and surfaces the findings, even though follow-up tracker creation remains a separate open runtime gap.
FEAT-002Running helix run does not discard unrelated worktree changes during recovery.tests/helix-cli.sh:2362SATISFIEDworkflows/TRACKER.md:232-250, scripts/helix:695-728, bash tests/helix-cli.sh passed on 2026-03-28.
FEAT-002Running helix backfill <scope> enforces the required trailers and durable report creation contract.tests/helix-cli.sh:2340-2341SATISFIEDscripts/helix:530-581, bash tests/helix-cli.sh passed on 2026-03-28.
TP-002The deterministic wrapper harness covers queue-drain control failure handling as well as successful NEXT_ACTION routing.noneUNTESTEDdocs/helix/03-test/test-plans/TP-002-helix-cli.md:26-39, scripts/helix:818-821, tests/helix-cli.sh:2330-2350. Current tests cover successful queue-drain outcomes but do not simulate run_check exiting non-zero before command substitution completes.
FEAT-002Running bash tests/helix-cli.sh remains the required deterministic verification path for wrapper behavior changes.docs/helix/03-test/test-plans/TP-002-helix-cli.md:12-16SATISFIEDworkflows/EXECUTION.md:194-212, bash tests/helix-cli.sh passed on 2026-03-28.

Gap Register

AreaClassificationPlanning EvidenceImplementation EvidenceResolution DirectionIssue
Queue-drain contract traceabilitySTALE_PLANdocs/helix/01-frame/features/FEAT-002-helix-cli.md:44-57, docs/helix/02-design/technical-designs/TD-002-helix-cli.md:75-90, workflows/EXECUTION.md:65-82scripts/helix:828-846, tests/helix-cli.sh:707-837plan-to-codehx-e7324e4f
Post-review follow-up creationINCOMPLETEdocs/helix/02-design/technical-designs/TD-002-helix-cli.md:186-192scripts/helix:623-631, scripts/helix:772-782, tests/helix-cli.sh:1662-1707code-to-planhx-c2dd330a
Queue-drain check-failure regression coverageINCOMPLETEdocs/helix/03-test/test-plans/TP-002-helix-cli.md:26-39, docs/helix/02-design/technical-designs/TD-002-helix-cli.md:75-101, workflows/EXECUTION.md:57-82scripts/helix:818-821, tests/helix-cli.sh:2330-2350code-to-planhx-7d00b06f
Tracker create help-path safetyINCOMPLETEworkflows/TRACKER.md:90-139, workflows/TRACKER.md:173-192scripts/tracker.sh:154-175, scripts/tracker.sh:792-796code-to-planhx-650ab26b
Test-plan evidence traceabilitySTALE_PLANdocs/helix/03-test/test-plans/TP-002-helix-cli.md:96-106, workflows/TRACKER.md:90-139tests/helix-cli.sh:1945-2014plan-to-codehx-4e5451cd
Build-plan tracker sequencingSTALE_PLANdocs/helix/04-build/implementation-plan.md:46-52, docs/helix/04-build/implementation-plan.md:70-74ddx bead list --status open --json on 2026-03-28plan-to-codehx-8447a41c
Packaging and installer contractALIGNEDworkflows/README.md:86-103, docs/helix/01-frame/prd.md:174-183, docs/helix/01-frame/features/FEAT-002-helix-cli.md:87-95scripts/install-local-skills.sh:10-45, scripts/install-local-skills.sh:92-110, .agents/skills/helix-align/SKILL.md:1-4, .agents/skills/helix-align -> ../../skills/helix-align, tests/helix-cli.sh:1071-1104, tests/validate-skills.sh:63-91code-to-planhx-3eb44502

Quality Findings

AreaDimensionConcernSeverityResolutionIssue
Tracker note historymaintainabilityRepeated implementation retries against unchanged non-execution issues append near-duplicate blocker notes, which makes the canonical tracker history harder to read.lowquality-improvementhx-bf99e0ee
Review evidence note historymaintainabilityRepeated repo-alignment refreshes append duplicate or near-duplicate verification notes to closed review records even when the proof lane and coverage state are unchanged.lowquality-improvementhx-44a5dbfe

Traceability Matrix

VisionRequirementFeature/StoryArch/ADRDesignTestsImpl PlanCode StatusClassification
Supervisory autopilot advances the weakest ready layerhelix run chooses the least-power next bounded actionFEAT-001 / US-001ADR-001SD-001, TD-002tests/helix-cli.sh:573-837docs/helix/04-build/implementation-plan.mdQueue-drain routing is implemented and testedALIGNED
Tracker-first executionTracker state is the durable execution layerFEAT-002 tracker modelADR-002API-001, TD-002tests/helix-cli.sh:1758-2058docs/helix/04-build/implementation-plan.mdCore tracker semantics are implemented; create-help safety still driftsINCOMPLETE
Authority-ordered reconciliationFunctionality changes route to align or plan before execution resumesFEAT-001 / US-001ADR-001SD-001, TD-002tests/helix-cli.sh:693-769docs/helix/04-build/implementation-plan.mdImplemented and testedALIGNED
Review happens before the loop advancesReview findings create or reopen follow-up workFEAT-002 review handlingADR-001TD-002tests/helix-cli.sh:1662-1707docs/helix/04-build/implementation-plan.mdLoop stops on findings, but follow-up tracker work is not created yetINCOMPLETE
Queue-drain failures fail closed without shell-state regressionsDeterministic tests should cover queue-drain check failure handlingTP-002 runtime queue controlADR-001TD-002tests/helix-cli.sh:2330-2350docs/helix/04-build/implementation-plan.mdRuntime guard exists in scripts/helix, but the failing branch is not yet covered by the harnessINCOMPLETE
HELIX ships as one package preserving skills/ plus workflows/Local installs preserve package-relative shared-resource accessFEAT-002 local installationADR-001SD-001tests/helix-cli.sh:1071-1104, tests/validate-skills.sh:63-91n/aInstaller and package surface remain alignedALIGNED
Test and build layers stay authoritativeTests and build sequencing cite current proof and live queue stateTP-002 and implementation plann/an/atests/helix-cli.sh, tests/validate-skills.shdocs/helix/04-build/implementation-plan.mdDocs exist but still cite stale evidence and superseded issue IDsSTALE_PLAN

Review Issue Summary

Review IssueFunctional AreaStatusKey FindingsRecommended Direction
hx-a3dab13bPlanning stack and durable report surfacescompleteFEAT-002, TP-002, and the 04-build implementation plan still contain stale traceability.plan-to-code
hx-5c8aee1eWrapper runtime and acceptance coveragecompleteQueue-drain routing and review surfacing remain implemented and verified; review-follow-up creation is still incomplete against TD-002, and the queue-drain check-failure regression path still lacks deterministic harness coverage.code-to-plan
hx-c41039f2Tracker queue hygiene and help safetycompleteCore tracker behavior remains aligned; ddx bead create --help still falls through to mutation parsing.code-to-plan
hx-3eb44502Packaging, installer, and validation surfacescompletePackaging and installer surfaces remain aligned with the contract; no new repo-scope packaging drift found.maintain

Execution Issues Generated

Issue IDTypeLabelsGoalDependenciesVerification
hx-e7324e4ftaskhelix,phase:design,kind:design,area:cli,feature:FEAT-002Update FEAT-002 so the feature spec matches the implemented PLAN and POLISH queue-drain contract.nonePending: docs/helix/01-frame/features/FEAT-002-helix-cli.md should align with TD-002, workflows/EXECUTION.md, scripts/helix, and tests/helix-cli.sh.
hx-c2dd330ataskhelix,phase:build,kind:review,area:cli,feature:FEAT-002Create or reopen follow-up tracker work when helix review finds issues.nonePending: deterministic coverage should prove tracker side effects and non-advancing loop behavior for non-clean review output.
hx-7d00b06ftaskhelix,phase:test,area:runtime,source:review,feature:FEAT-002Add deterministic regression coverage for queue-drain check failure after ready work exhausts.nonePending: tests/helix-cli.sh should assert helix: check failed after queue drain and absence of a set -u unbound-variable failure when run_check exits non-zero before emitting NEXT_ACTION.
hx-650ab26bbughelix,phase:build,kind:review,area:tracker,source:reviewTreat ddx bead create --help and -h as usage paths instead of tracker mutations.nonePending: deterministic coverage should prove help output prints usage and appends no tracker record.
hx-8447a41ctaskhelix,phase:build,area:docs,source:review,feature:FEAT-002Refresh docs/helix/04-build/implementation-plan.md so it cites the live execution or refinement issues instead of superseded predecessors.nonePending: docs/helix/04-build/implementation-plan.md should sequence the current tracker-backed slices only.
hx-4e5451cdtaskhelix,phase:test,area:docs,source:review,feature:FEAT-002Refresh TP-002 evidence references so they cite authoritative contract or deterministic harness sources.nonePending: docs/helix/03-test/test-plans/TP-002-helix-cli.md should cite current supporting evidence only.
hx-bf99e0eetaskhelix,phase:iterate,kind:backlog,area:tracker,source:reviewDeduplicate repeated implementation blocker notes for unchanged non-execution issues.nonePending: deterministic coverage should prove identical retries no longer append near-duplicate blocker notes.
hx-44a5dbfetaskhelix,phase:iterate,kind:backlog,area:tracker,source:reviewDeduplicate repeated review evidence notes during alignment refreshes so same-state reruns do not bloat closed review records.nonePending: deterministic coverage should prove unchanged same-scope alignment refreshes do not append duplicate review evidence notes.

This execution reused the canonical closed review epic hx-1297bf62 and its four functional-area review tasks hx-a3dab13b, hx-5c8aee1e, hx-c41039f2, and hx-3eb44502. Revalidation against the live planning stack, wrapper runtime, tracker, installer, and proof lane found one additional uncovered repo-scope gap since the prior refresh: runtime test issue hx-7d00b06f for queue-drain check-failure coverage. No new review epic or review task was created; the existing canonical review container was updated in place and the durable report was refreshed against the live eight-issue open set.

Issue Coverage Verification

Gap / CriterionCovering IssueStatus
FEAT-002 stale queue-drain contracthx-e7324e4fcovered
Review findings do not yet create or reopen follow-up tracker workhx-c2dd330acovered
Queue-drain check failure path lacks deterministic regression coveragehx-7d00b06fcovered
ddx bead create --help mutates tracker statehx-650ab26bcovered
docs/helix/04-build/implementation-plan.md still cites superseded issue IDshx-8447a41ccovered
TP-002 stale evidence referenceshx-4e5451cdcovered
Repeated identical implementation blocker notes append duplicate historyhx-bf99e0eecovered
Repeated identical alignment-refresh notes append duplicate review evidence historyhx-44a5dbfecovered

Execution Order

  1. hx-e7324e4f to restore feature-spec authority before more wrapper/runtime drift accumulates.
  2. hx-c2dd330a to complete the remaining review-handling behavior required by TD-002.
  3. hx-7d00b06f to lock in regression coverage for the queue-drain check-failure guard before future refactors erode the fix.
  4. hx-650ab26b to make tracker help requests non-mutating.
  5. hx-8447a41c to refresh the 04-build layer so it cites the live queue accurately.
  6. hx-4e5451cd to repair the stale TP-002 evidence references.
  7. hx-bf99e0ee as backlog quality work on implementation-note hygiene.
  8. hx-44a5dbfe as backlog quality work on alignment-review note hygiene.

Critical Path: hx-e7324e4f then hx-c2dd330a Parallel: hx-7d00b06f, hx-650ab26b, hx-8447a41c, hx-4e5451cd, hx-bf99e0ee, hx-44a5dbfe Blockers: none First Recommended Execution Set: execution-safe now: hx-c2dd330a, hx-650ab26b, hx-44a5dbfe, and hx-7d00b06f; design-led prerequisite outside the execution-safe queue: hx-e7324e4f

Open Decisions

DecisionWhy OpenGoverning ArtifactsRecommended Owner
When helix review reports findings, should the wrapper reopen the implementation issue, create a new follow-up issue, or choose between them based on machine-readable trailers?TD-002 requires follow-up tracker work before the loop advances, but the exact tracker mutation policy is still not explicit enough to implement without interpretation.docs/helix/01-frame/features/FEAT-002-helix-cli.md, docs/helix/02-design/technical-designs/TD-002-helix-cli.md, docs/helix/04-build/implementation-plan.mdHELIX maintainers

Queue Health and Exhaustion Assessment

  • Actionable follow-up work remains for repo scope; the queue is not exhausted.
  • This pass reused the closed canonical repo review epic hx-1297bf62 and its area tasks hx-a3dab13b, hx-5c8aee1e, hx-c41039f2, and hx-3eb44502 rather than creating another same-scope review tree.
  • Fresh verification in this execution pass succeeded: bash tests/helix-cli.sh (PASS: 77 helix wrapper tests), bash tests/validate-skills.sh (validated 10 HELIX skills), and git diff --check.
  • Current tracker health is 165 total issues, 8 open, 0 in-progress, and 156 closed. ddx bead list --status open --json confirms the remaining eight open issues are exactly hx-e7324e4f, hx-c2dd330a, hx-4e5451cd, hx-8447a41c, hx-bf99e0ee, hx-650ab26b, hx-44a5dbfe, and hx-7d00b06f. The remaining non-open, non-closed record stays deferred historical artifact hx-e4d48132, confirmed by ddx bead list --status deferred --json.
  • ddx bead ready --json --execution now returns hx-c2dd330a, hx-650ab26b, hx-44a5dbfe, and hx-7d00b06f. The design and documentation follow-up issues remain intentionally open but not execution-safe.
  • This alignment pass corrected hx-7d00b06f from review-only metadata to an actual test execution issue by updating it to phase:test, spec-id: TP-002, and execution-eligible: true, because the real remaining work is deterministic harness coverage rather than more review bookkeeping.
  • Revalidation against the current planning stack, runtime, tracker, installer, and proof lane found no additional repo-scope gaps beyond hx-e7324e4f, hx-c2dd330a, hx-7d00b06f, hx-650ab26b, hx-8447a41c, hx-4e5451cd, hx-bf99e0ee, and hx-44a5dbfe.
  • Repo scope remains ALIGNED for packaging, installer behavior, queue-drain dispatch, tracker core semantics, and deterministic proof-lane stability. The remaining non-aligned areas are the previously tracked stale-plan and incomplete gaps plus the newly recorded runtime regression-coverage gap above.