All notable changes to Nova Agent Workflows should be recorded here.
Nova Serious Workflow v1.8 public-safe governance update.
Added:
- context contract before meaningful edits
- Goal/Rider task framing for serious
T###work - structured risk approval pattern for risky actions
- runtime harness action flow: context -> allowed surfaces -> proposed action -> verdict -> execution -> evidence -> sync-back
- verdict vocabulary:
allow,warn,review,block - retry budget rule for repeated same-root-cause failures
- task-thread event shape for long-running, handoff-heavy, or multi-agent work
- agentic delivery chain: clarify/domain model -> PRD or compact spec -> vertical-slice issue plan -> optional TDD -> architecture review
Safety:
- no private memory export
- no runtime logs
- no private machine paths
- no hooks, MCP registration, installers, package scripts, account actions, deploys, or secret handling added
- no tagged release, package publication, deployment, force-push, remote rewrite, or repository setting change performed for v0.3.0-draft
- push to GitHub remains blocked until exact user approval after fresh scrub and smoke review
Nova Serious Workflow v1.7 planner/orchestrator/review overlay.
Added:
- public-safe planner/orchestrator/review overlay for serious app, feature, GitHub, and PR work
- assumptions-first and simple-first guidance
- new-app planning pattern: concise project docs plus approved issue plan when useful
- existing-feature planning pattern: small dependency-ordered tasks or issues
- implementation delivery pattern: one approved issue -> one branch or PR -> checks -> review/fix -> user QA handoff
- outside-review pattern: question whether the change should exist, whether a simpler path exists, and whether the real code path supports the diff
- clarification pattern: ask one plain-language question when the goal is blurry
- board-finisher GitHub delivery guard for remote actions
Safety:
- no private memory export
- no runtime logs
- no private machine paths
- no archived skill code copied into the public pack
- no GitHub issue creation automation added
- no tagged release, package publication, deployment, force-push, or remote rewrite performed for v0.2.2-draft
Nova Serious Workflow v1.6 mini-plan alignment.
Updated:
- required a visible six-layer mini-plan before serious
T###work - aligned the six layers to:
- Obsidian Brain Layer
- Planning / Spec Layer
- GitHub Delivery Layer
- Verification / Risk Gate Layer
- Release / Publish Layer
- Sync Back / Learning Layer
- clarified that unused layers should be marked
N/Ainstead of inventing GitHub, release, or spec work - added a legacy project alignment rule: do not rewrite old completed tasks just to retrofit the new format
- updated Codex, Claude Code, board-finisher, and project guidance wording
Safety:
- no private memory export
- no runtime logs
- no private machine paths
- no tagged release, package publication, deployment, force-push, or remote rewrite performed for v0.2.1-draft
Nova Serious Workflow v1.5 alignment.
Added:
- six-layer serious workflow stack:
- Brain / context layer
- Ingestion / knowledge layer
- Planning / spec layer
- Task board / checklist layer
- Verification / risk gate layer
- Delivery / sync-back layer
- planning templates:
templates/PRD.mdtemplates/SPEC.mdtemplates/ARCHITECTURE.mdtemplates/AGENTS.md
- Codex and Claude Code adapter wording for the six-layer workflow
- release-readiness update for the existing public repository state
Safety:
- no private memory export
- no runtime logs
- no private machine paths
- v0.2.0-draft was pushed to
mainafter local scrub - no tagged release, package publication, deployment, or remote rewrite performed for v0.2.0-draft
- release gate remains required before any tagged release or future risky public action
Initial local draft.
Added:
- core workflow:
nova-serious-workflow - core workflow:
nova-debug-review-gate - core workflow:
nova-board-finisher - Codex skill adapters
- Claude Code skill adapters
CLAUDE.md.example- task board, checklist, work log, and handoff templates
- synthetic game-project and Obsidian-brain examples
- public export scrub checklist
- risk gates
- versioning policy
Safety:
- no private memory export
- no runtime logs
- no account, deploy, payment, secret, browser-account, or OS repair automation
- no hooks that run by default
- no public repository created
- no publish or push performed
Smoke review update:
- clarified all Codex and Claude Code adapter skills so they can operate safely even when copied without the full
workflows/folder - added T048 scrub report
Version routine update:
- added Nova alignment rule to
VERSIONING.md - added stable private routine note reference for future workflow-pack upgrades
- updated Nova Brain Evolution Dashboard privately as Ver 1.4*
Permanent release gate update:
- added
safety/PRE_PUBLISH_RELEASE_GATE.md - added
RELEASE_READINESS.md - added
.gitignore - made fresh scrub plus cross-agent smoke review mandatory before any public release
- clarified that GitHub repository creation, remote setup, push, publish, and release require explicit approval for the exact action
- added
safety/PUBLIC_EXPORT_SCRUB_REPORT_T051.md
Release prep update:
- finalized MIT license copyright as
2026 Nova Agent Workflows - release remains gated on final scrub, local commit, GitHub repository creation, and push verification
- added
safety/PUBLIC_EXPORT_SCRUB_REPORT_T052.md