@tiic-tech/openworkflow 0.1.1 → 0.1.2

package/dist/core/src/commands/registry.js

@@ -9,20 +9,30 @@ export const WORKFLOW_COMMANDS = [
         ".openworkflow/context/CONTEXT_MAP.yaml",
     ]),
     command("vision", ["build-vision"], "Create or refine the product vision contract through focused collaboration.", "vision", [".openworkflow/vision/VISION.md", ".openworkflow/vision/VISION_CONTRACT.yaml"], visionProtocol()),
-    command("validation", ["build-validation"], "Prioritize the core feature or assumption that must be validated first.", "validation", [".openworkflow/validation/"], validationProtocol()),
-    command("proto", ["build-prototype", "ow:prototype"], "Build the smallest prototype needed to validate the current core feature.", "prototype", [".openworkflow/prototypes/"], prototypeProtocol()),
-    command("tune", ["ow:tune:proto"], "Revise the current prototype and record the decision audit automatically.", "prototype", [".openworkflow/prototypes/", ".openworkflow/decisions/"], tuneProtocol()),
+    command("validation", ["build-validation"], "Compile proto-ready vision into one prototype validation target.", "validation", [".openworkflow/validation/"], validationProtocol()),
+    command("build-proto-prompt", ["build-prototype-prompt"], "Internally compile ready vision and validation into ready prototype prompt packs.", "prototype", [".openworkflow/prototypes/"], buildProtoPromptProtocol(), "internal"),
+    command("vision2prompt", ["vision-to-strategic-prototype-prompt"], "Internally compile ready vision and validation into strategic prototype prompt text.", "prototype", [".openworkflow/prototypes/"], vision2PromptProtocol(), "internal"),
+    command("prompt2proto", ["prompt-to-prototype-images"], "Internally generate high-fidelity prototype image groups from prepared prompt text.", "prototype", [".openworkflow/prototypes/"], prompt2ProtoProtocol(), "internal"),
+    command("proto", ["build-prototype", "ow:prototype"], "Create image-first strategic prototype prompt packs from vision or validation context.", "prototype", [".openworkflow/prototypes/"], prototypeProtocol()),
+    command("tune", ["tune-prototype", "ow:tune:proto"], "Refine accepted prototype screens or prompt packs and record the decision audit automatically.", "prototype", [".openworkflow/prototypes/", ".openworkflow/decisions/"], tuneProtocol()),
     command("decision", ["build-decision"], "Internally record prototype review outcomes for audit.", "decision", [".openworkflow/decisions/"], decisionProtocol(), "internal"),
     command("design", ["build-design"], "Convert accepted prototype evidence into product design for production specification.", "design", [".openworkflow/design/"], designProtocol()),
-    command("spec", ["build-spec"], "Create one focused production spec from accepted product design.", "spec", [
-        ".openworkflow/specs/",
-    ]),
-    command("change", ["build-change"], "Create one focused production change for the current core feature.", "change", [
-        ".openworkflow/changes/",
-    ]),
-    command("team", ["run-team", "build-team"], "Execute approved production work through the Agent Team runtime.", "runtime", [
-        ".openworkflow/runtime/",
-    ]),
+    command("spec", ["build-spec"], "Create one focused production spec from accepted product design.", "spec", [".openworkflow/specs/"], specProtocol()),
+    command("change", ["build-change"], "Create one focused production change for the current core feature.", "change", [".openworkflow/changes/"], changeProtocol()),
+    command("team", ["run-team", "build-team"], "Execute approved production work through the Agent Team runtime.", "runtime", [".openworkflow/runtime/"], teamProtocol()),
+    command("decompose-to-changes", [], "Create, update, query, or maintain an OpenWorkflow candidate change queue.", "planning", [
+        "changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+        "changes/<plan_id>/CANDIDATE_CHANGES.md",
+        "changes/<plan_id>/SUMMARY.yaml",
+    ], decomposeToChangesProtocol()),
+    command("analyze-changes", [], "Analyze multiple candidate change queues and recommend the next queue and candidate without selecting it.", "planning", ["changes/<analysis_id>/CHANGE_ANALYSIS.yaml", "changes/<analysis_id>/CHANGE_ANALYSIS.md"], analyzeChangesProtocol()),
+    command("select-change", [], "Select one implementable candidate change and create implementation-ready planning artifacts.", "planning", [
+        "changes/<plan_id>/<candidate-id>-<slug>/SELECTED_CHANGE.yaml",
+        "changes/<plan_id>/<candidate-id>-<slug>/ATOM_TASKS.yaml",
+        "changes/<plan_id>/<candidate-id>-<slug>/IMPLEMENTATION_BRIEF.md",
+    ], selectChangeProtocol()),
+    command("git-automation", [], "Operate the managed git lifecycle shell for local branch, commit, PR-ready summary, and remote approval gates.", "governance", ["changes/<plan_id>/CANDIDATE_CHANGES.yaml", "changes/<plan_id>/PR_READY_SUMMARY.md"], gitAutomationProtocol()),
+    command("coder", ["code-quality-governor", "engineering-quality"], "Govern source edits with internal Agent-only code quality preflight, RED/GREEN evidence, self-check, validation, and evidence binding.", "execution", [], coderProtocol(), "internal"),
 ];
 export function getWorkflowCommands() {
     return WORKFLOW_COMMANDS;
@@ -40,10 +50,409 @@ function command(id, legacyIds, description, stage, targetArtifacts, protocol, v
         protocol,
     };
 }
+function decomposeToChangesProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "candidate-queue-decomposition-and-maintenance",
+        requiredContext: [
+            "references/planning-artifact-contracts.md",
+            "skills/decompose-to-changes/references/decomposition-protocol.md",
+        ],
+        optionalContext: [
+            "changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+            "changes/<plan_id>/SUMMARY.yaml",
+            "changes/<plan_id>/HIGH_RISK_DECISION_REPORT.md",
+            "docs/OW_DEVELOP_PLAN.md",
+            "docs/OW_DEVELOP_PLAN_Phase2.md",
+            "user-provided planning source",
+        ],
+        forbiddenContext: [],
+        allowedOutputs: [
+            "changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+            "changes/<plan_id>/CANDIDATE_CHANGES.md",
+            "changes/<plan_id>/SUMMARY.yaml",
+            "changes/<plan_id>/HIGH_RISK_DECISION_REPORT.md when the next actionable work is high risk",
+        ],
+        conditionalOutputs: [
+            "high-risk decision report when a risk: high candidate becomes the next actionable work",
+            "queue maintenance operation entries for add, update, split, merge, defer, block, supersede, restore, or complete",
+        ],
+        forbiddenOutputs: [
+            "SELECTED_CHANGE.yaml",
+            "ATOM_TASKS.yaml",
+            "IMPLEMENTATION_BRIEF.md",
+            "implementation code changes",
+            "generated .agents/** or .openworkflow/** edits unless selected and explicitly approved",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Run git status --short --branch and record branch and dirty-tree state.",
+                "Decide whether this is new decomposition or maintenance of an existing queue.",
+                "Run the queue scope gate: choose one feature, bounded module, command surface, artifact family, or workflow slice for this queue.",
+                "Read existing queue YAML before changing candidate ids or statuses.",
+            ],
+            during: [
+                "For every new CANDIDATE_CHANGES queue, set queue_policy.git_lifecycle_gate: strict and a plan-owned queue_policy.branch_boundary before selecting work.",
+                "Preserve stable candidate ids, git_lifecycle_gate, and branch_boundary when updating an existing queue.",
+                "Record features outside the current queue boundary as deferred refs instead of current candidates.",
+                "Keep candidates focused, dependency-aware, and bounded by owned paths.",
+                "Append an operation entry for every queue maintenance edit.",
+            ],
+            after: [
+                "Refresh CANDIDATE_CHANGES.md as a readable view of YAML source truth.",
+                "Refresh SUMMARY.yaml with candidate count, next recommended candidate, risks, and validation evidence.",
+                "Do not mark a queue completed until PR evidence is recorded through DRAFT_PR_OPERATION_EVIDENCE.yaml or the queue is explicitly blocked before PR creation.",
+                "Stop with a high-risk report instead of selecting or implementing risk: high candidates.",
+            ],
+        },
+        antiPatterns: [
+            "Do not select a candidate from decompose-to-changes.",
+            "Do not implement code from decompose-to-changes.",
+            "Do not create a new top-level changes folder for every small candidate inside the same feat boundary.",
+            "Do not turn one CANDIDATE_CHANGES queue into a roadmap bucket for multiple features or a large module family.",
+            "Do not delete historical candidate ids; use status transitions and operation evidence.",
+        ],
+        handoffCommands: ["/ow:analyze-changes", "/ow:select-change"],
+    };
+}
+function analyzeChangesProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "read-only-cross-queue-priority-analysis",
+        requiredContext: [
+            "references/planning-artifact-contracts.md",
+            "skills/analyze-changes/references/analysis-protocol.md",
+            "changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+        ],
+        optionalContext: [
+            "references/git-version-control-governance.md",
+            "references/issue-governance.md",
+            "changes/*/CANDIDATE_CHANGES.yaml when the user asks for global or cross-queue analysis",
+            "changes/<plan_id>/SUMMARY.yaml",
+            "changes/<plan_id>/HIGH_RISK_DECISION_REPORT.md",
+        ],
+        forbiddenContext: [],
+        allowedOutputs: [
+            "changes/<analysis_id>/CHANGE_ANALYSIS.yaml",
+            "changes/<analysis_id>/CHANGE_ANALYSIS.md",
+        ],
+        conditionalOutputs: [
+            "high-risk stop recommendation that points to the needed HIGH_RISK_DECISION_REPORT.md",
+            "queue maintenance recommendation when no candidate is safe to select",
+        ],
+        forbiddenOutputs: [
+            "CANDIDATE_CHANGES.yaml mutations unless the user separately requests maintenance",
+            "SELECTED_CHANGE.yaml",
+            "ATOM_TASKS.yaml",
+            "IMPLEMENTATION_BRIEF.md",
+            "implementation code changes",
+            "high-risk implementation approval",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Run git status --short --branch and record branch and dirty-tree state.",
+                "Confirm this is a cross-queue decision; when only one queue is active, hand off to select-change.",
+                "Discover only user-provided queues, or obvious changes/*/CANDIDATE_CHANGES.yaml files when global analysis is requested.",
+                "Read YAML queues as source truth and Markdown views only as aids.",
+            ],
+            during: [
+                "Score candidates by readiness, dependency unlock value, risk, branch fit, dirty-tree fit, Issue linkage, validation realism, and user recency.",
+                "Treat high-risk candidates as stop recommendations unless a concrete high-risk option is already approved.",
+                "Recommend exactly one target plan id and candidate id only when evidence supports selection.",
+            ],
+            after: [
+                "Write CHANGE_ANALYSIS.yaml before CHANGE_ANALYSIS.md.",
+                "Record rejected alternatives with plan id, candidate id, and concise reasons.",
+                "Hand off to select-change without mutating selection artifacts.",
+            ],
+        },
+        antiPatterns: [
+            "Do not select candidates from analyze-changes.",
+            "Do not implement candidates from analyze-changes.",
+            "Do not treat CHANGE_ANALYSIS.yaml as approval for high-risk implementation.",
+            "Do not discover every queue unless the user requests global comparison.",
+            "Do not use analyze-changes as a mandatory pre-step for single-queue selection.",
+        ],
+        handoffCommands: ["/ow:select-change", "/ow:decompose-to-changes"],
+    };
+}
+function selectChangeProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "single-candidate-selection-and-atomization",
+        requiredContext: [
+            "references/planning-artifact-contracts.md",
+            "skills/select-change/references/selection-protocol.md",
+            "changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+        ],
+        optionalContext: [
+            "changes/<plan_id>/CANDIDATE_CHANGES.md",
+            "changes/<analysis_id>/CHANGE_ANALYSIS.yaml only for cross-queue recommendations",
+            "changes/<plan_id>/SUMMARY.yaml",
+            "changes/<plan_id>/HIGH_RISK_DECISION_REPORT.md",
+        ],
+        forbiddenContext: [],
+        allowedOutputs: [
+            "changes/<plan_id>/<candidate-id>-<slug>/SELECTED_CHANGE.yaml",
+            "changes/<plan_id>/<candidate-id>-<slug>/ATOM_TASKS.yaml",
+            "changes/<plan_id>/<candidate-id>-<slug>/IMPLEMENTATION_BRIEF.md",
+            "selection and operation entries in changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+            "refreshed changes/<plan_id>/CANDIDATE_CHANGES.md",
+        ],
+        conditionalOutputs: [
+            "rejected alternatives copied from CHANGE_ANALYSIS.yaml when consuming cross-queue analysis",
+            "targeted readiness report when the user asks to inspect a candidate without selecting it",
+        ],
+        forbiddenOutputs: [
+            "implementation code changes",
+            "local commits, stashes, resets, branch switches, or destructive git operations",
+            "risk: high selection without explicit approval of a concrete option from HIGH_RISK_DECISION_REPORT.md",
+            "generated .agents/** or .openworkflow/** edits unless selected and explicitly approved",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Run git status --short --branch and compare current branch with queue_policy.branch_boundary.",
+                "Check dirty-tree state and stop if unrelated work would contaminate the selected change.",
+                "For a single active queue, rank candidates directly without requiring analyze-changes.",
+                "Confirm candidate dependencies, readiness, risk, owned paths, validation, and acceptance.",
+            ],
+            during: [
+                "Use next_recommended_candidate_id, dependency satisfaction, unlock value, selection_policy, risk, owned paths, and validation realism to choose inside one queue.",
+                "Select exactly one candidate inside the owning queue folder.",
+                "Re-check high-risk approval before writing selection artifacts for risk: high candidates.",
+                "Keep atom tasks small enough for one focused implementation pass.",
+            ],
+            after: [
+                "Update candidate status to selected and append a selection operation.",
+                "Refresh the readable Markdown queue view.",
+                "Stop before implementation unless the user explicitly asks to continue.",
+            ],
+        },
+        antiPatterns: [
+            "Do not silently select a high-risk candidate.",
+            "Do not select on the wrong branch without an explicit planning-only exception.",
+            "Do not mark the candidate done from select-change.",
+            "Do not create a new top-level changes folder for a candidate inside an existing feat queue.",
+        ],
+        handoffCommands: ["/ow:change", "/ow:team", "/ow:git-automation"],
+    };
+}
+function gitAutomationProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "managed-git-lifecycle-shell",
+        requiredContext: [
+            "references/git-version-control-governance.md",
+            "references/gh-operation-governance.md",
+            "changes/<plan_id>/CANDIDATE_CHANGES.yaml",
+        ],
+        optionalContext: [
+            "changes/<plan_id>/HIGH_RISK_DECISION_REPORT.md",
+            "changes/<plan_id>/PR_READY_SUMMARY.md",
+            "changes/<plan_id>/<candidate-id>-<slug>/LOCAL_COMMIT_EVIDENCE.yaml",
+        ],
+        forbiddenContext: [],
+        allowedOutputs: [
+            "local branch checkout or creation through openworkflow git-automation branch",
+            "local selected-change commit through openworkflow git-automation commit",
+            "local PR_READY_SUMMARY.md through openworkflow git-automation summary",
+            "remote operation plan through openworkflow git-automation remote",
+            "remote read-only PR-ready plan through openworkflow git-automation remote-plan",
+            "draft PR pilot preview or explicitly gated mutation through openworkflow git-automation draft-pr",
+            "local evidence artifacts under changes/<plan_id>/",
+        ],
+        conditionalOutputs: [
+            "high-risk decision report when remote mutation or autonomous mode is requested",
+            "follow-up CANDIDATE_CHANGES entry for autonomous git automation",
+        ],
+        forbiddenOutputs: [
+            "git push without explicit operation-level user approval",
+            "gh pr create/edit/merge without explicit operation-level user approval",
+            "gh issue create/edit/close without explicit operation-level user approval",
+            "git reset, rebase, force-push, or destructive branch deletion",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Read the queue branch boundary and confirm current git state.",
+                "Confirm whether the requested mode is managed or autonomous.",
+                "Stop on autonomous or remote mutation requests unless a high-risk approval exists for exact operations.",
+            ],
+            during: [
+                "Use dry-run or preview before any local mutation.",
+                "Record plan id, candidate id, branch, dirty paths, command preview, validation evidence, and affected paths.",
+                "Keep local branch, commit, and summary actions scoped to the selected queue.",
+                "When committing implementation work, write LOCAL_COMMIT_EVIDENCE.yaml under the selected-change folder and reference it from completion evidence.",
+            ],
+            after: [
+                "Record commit hash and evidence path when a local commit is created.",
+                "Confirm the selected candidate completion records implementation_changed_files: true and a repo-relative LOCAL_COMMIT_EVIDENCE.yaml path.",
+                "Regenerate PR_READY_SUMMARY.md after commit evidence changes when appropriate.",
+                "Report remote operations as gated and include ordered commit or queue evidence for push, PR, and merge planning.",
+            ],
+        },
+        antiPatterns: [
+            "Do not treat git-automation enabled as permission to push, merge, or mutate GitHub in this G015 shell.",
+            "Do not hide dirty paths or omit command previews from evidence.",
+            "Do not create a selected change with no local commit when implementation changed files.",
+            "Do not batch multiple completed selected changes into one checkpoint commit to satisfy commit evidence.",
+            "Do not amend only to force a commit to contain its own hash.",
+        ],
+        handoffCommands: [
+            "openworkflow git-automation branch --root . --queue changes/<plan_id>/CANDIDATE_CHANGES.yaml --json",
+            "openworkflow git-automation commit --root . --queue changes/<plan_id>/CANDIDATE_CHANGES.yaml --candidate <id> --message <msg> --validation-evidence <cmds> --commit-evidence --json",
+            "openworkflow git-automation summary --root . --queue changes/<plan_id>/CANDIDATE_CHANGES.yaml --json",
+            "openworkflow git-automation simulate --root . --queue changes/<plan_id>/CANDIDATE_CHANGES.yaml --base <base-ref> --json",
+            "openworkflow git-automation remote-plan --root . --queue changes/<plan_id>/CANDIDATE_CHANGES.yaml --base <base-ref> --remote <remote> --target-base <branch> --json",
+            "openworkflow git-automation draft-pr --root . --queue changes/<plan_id>/CANDIDATE_CHANGES.yaml --base <base-ref> --remote <remote> --target-base <branch> --json",
+        ],
+        internalSections: [
+            {
+                tag: "mode_policy",
+                items: [
+                    "managed mode may perform approved local branch, commit, and summary operations with previews and evidence.",
+                    "managed mode must gate remote push, PR, Issue, and merge operations behind explicit user approval while producing a clear operation plan.",
+                    "remote-plan mode may read remote refs and PR metadata, but must not push, create PRs, edit PRs, merge, or mutate Issues.",
+                    "draft-pr mode is disabled by default; mutation requires --write, --allow-draft-pr, current remote-plan evidence, and rollback guidance.",
+                    "autonomous mode is a future high-risk path and is not implemented by the G015 command shell.",
+                ],
+            },
+            {
+                tag: "evidence_policy",
+                items: [
+                    "Every git operation must be traceable to a plan id, candidate id, command preview, before and after state, and validation evidence when applicable.",
+                    "Remote approval handoff must include branch, target base, ordered local commits, PR-ready summary path, conflict-resolution checkpoint, and merge evidence expectations.",
+                    "A selected change must have at least one local commit when implementation changed files, recorded in changes/<plan_id>/<candidate-id>-<slug>/LOCAL_COMMIT_EVIDENCE.yaml.",
+                    "The owning queue completion must reference the selected-change LOCAL_COMMIT_EVIDENCE.yaml path; manual checkpoint batching is a failure mode, not valid evidence.",
+                    "Follow-up evidence commits are allowed when they preserve the selected-change HEAD relationship.",
+                ],
+            },
+        ],
+    };
+}
+function coderProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "internal-agent-code-execution-governance",
+        requiredContext: [
+            "references/internal-coder-protocol.md",
+            "skills/coder/SKILL.md",
+            "changes/<plan_id>/CANDIDATE_CHANGES.yaml when queue-driven",
+            "changes/<plan_id>/<candidate-id>-<slug>/SELECTED_CHANGE.yaml when selected-change-driven",
+            "git status --short --branch",
+        ],
+        optionalContext: [
+            "references/validation-trust-domains.md",
+            "references/skill-system-lifecycle.md",
+            "references/git-version-control-governance.md",
+            "changes/<plan_id>/<candidate-id>-<slug>/ATOM_TASKS.yaml",
+            "changes/<plan_id>/<candidate-id>-<slug>/IMPLEMENTATION_BRIEF.md",
+            "changes/<plan_id>/<candidate-id>-<slug>/LOCAL_COMMIT_EVIDENCE.yaml",
+        ],
+        forbiddenContext: [
+            ".openworkflow/runtime/** unless the active task is /ow:team execution",
+            "generated .agents/** as source truth",
+        ],
+        allowedOutputs: [
+            "source edits already allowed by the current selected change",
+            "test, fixture, verifier, or reference edits already allowed by the current selected change",
+            "selected-change evidence updates under changes/<plan_id>/<candidate-id>-<slug>/",
+            "source-driven generated surfaces only after sync when the selected change owns them",
+        ],
+        conditionalOutputs: [
+            "LOCAL_COMMIT_EVIDENCE.yaml through openworkflow git-automation commit when implementation files changed",
+            "future optional CODER_EVIDENCE.yaml only after a later selected change defines the artifact contract",
+        ],
+        forbiddenOutputs: [
+            "user-facing command handoff to /ow:coder",
+            "new mandatory CODER_EVIDENCE.yaml",
+            "generated .agents/** hand edits",
+            ".openworkflow/** mutation unless the active user-facing command owns that output",
+            "git push, PR creation, Issue mutation, merge, reset, rebase, force-push, or destructive branch operations",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Recover trust with repo-local resume, handoff, inspect --strict, and git status before source edits.",
+                "Identify selected plan id, candidate id, selected-change artifact, owned paths, forbidden paths, and validation commands when queue-driven.",
+                "Build an owner/file/dependency map that names source truth, derived surfaces, validators, tests or fixtures, public report surfaces, and docs or skills.",
+                "Decide whether RED evidence is required for the change type; for docs-only or contract-only work, record why RED is not applicable.",
+            ],
+            during: [
+                "Change the source owner first and regenerate derived surfaces only through the repo-local sync path when generated surfaces are in scope.",
+                "For behavior, validator, CLI report, generated-surface, path-safety, summary, queue, or git-evidence changes, produce RED evidence before production edits when practical.",
+                "After edits, rerun the RED evidence or nearest equivalent and record GREEN evidence for the touched trust domain.",
+                "Keep /ow:change responsible for selected-change boundaries and /ow:team responsible for managed execution; coder governance constrains source edits but does not replace either command.",
+            ],
+            after: [
+                "Run the post-write self-check from skills/coder/SKILL.md before final validation.",
+                "Run the narrowest honest validation ladder for the touched trust domain.",
+                "Bind validation evidence to SELECTED_CHANGE.yaml, ATOM_TASKS.yaml, IMPLEMENTATION_BRIEF.md, or LOCAL_COMMIT_EVIDENCE.yaml as appropriate.",
+                "Report generated surfaces as source-driven or intentionally untouched; do not present /ow:coder as a normal user-facing next step.",
+            ],
+        },
+        antiPatterns: [
+            "Do not expose /ow:coder as a normal user-facing workflow entrypoint.",
+            "Do not use /ow:coder to bypass /ow:change selected-change boundaries or /ow:team execution governance.",
+            "Do not patch generated .agents surfaces as the durable fix.",
+            "Do not require CODER_EVIDENCE.yaml before a later candidate defines and proves the evidence contract.",
+            "Do not treat npm run build alone as sufficient validation for behavior, validator, generated-surface, summary, or git-evidence changes.",
+            "Do not hide historical validation debt; classify it separately from active-change failures.",
+        ],
+        internalSections: [
+            {
+                tag: "internal_command_boundary",
+                items: [
+                    "/ow:coder is internal and Agent-only.",
+                    "It is not a normal user-facing coding command and must not appear as the recommended handoff for product work.",
+                    "It constrains source-edit quality inside /ow:change, /ow:team, git-automation, and other implementation flows.",
+                    "Its source behavior lives in skills/coder/SKILL.md and references/internal-coder-protocol.md.",
+                ],
+            },
+            {
+                tag: "preflight_owner_map",
+                items: [
+                    "Identify source truth before edits: command registry, artifact registry, schema, validator, adapter template, source skill, or planning queue.",
+                    "Identify derived surfaces such as .agents/**, .openworkflow/audit/**, summaries, fixtures, and readable Markdown views.",
+                    "Name forbidden paths and dependency order before writing across multiple concerns.",
+                    "If two owners define the same rule, either collapse the duplication or record a temporary compatibility boundary.",
+                ],
+            },
+            {
+                tag: "red_green_evidence",
+                items: [
+                    "For behavior, validator, CLI report, generated-surface, path-safety, summary, queue, or git-evidence changes, prefer RED evidence before production edits.",
+                    "Valid RED evidence can be a failing parsed JSON/YAML assertion, fixture, generated-surface parity check, summary or resume health assertion, compile failure, or runtime verifier failure.",
+                    "After edits, rerun the RED evidence or nearest equivalent and record GREEN evidence.",
+                    "For docs-only, contract-only, mechanical rename, or exploratory work, mark RED not applicable and use the nearest structural check.",
+                ],
+            },
+            {
+                tag: "validation_ladder",
+                items: [
+                    "Use the narrowest honest validation ladder for the touched trust domain.",
+                    "Source skill or generated protocol work requires build, sync, generated diff review, strict inspect, and diff check.",
+                    "Command registry or generated runtime surface work requires build, sync, runtime-surface verification, and generated diff review.",
+                    "Artifact, schema, or validator work requires build, validate, and targeted valid/invalid fixtures.",
+                    "Git evidence work requires git-automation preview/write plus strict read-model commands.",
+                ],
+            },
+            {
+                tag: "evidence_binding",
+                items: [
+                    "Bind scope and acceptance to SELECTED_CHANGE.yaml.",
+                    "Bind task status and verification results to ATOM_TASKS.yaml.",
+                    "Bind handoff instructions to IMPLEMENTATION_BRIEF.md.",
+                    "Bind local commit hashes and validation evidence to LOCAL_COMMIT_EVIDENCE.yaml when implementation files changed.",
+                    "Do not batch multiple completed selected changes into one checkpoint commit.",
+                ],
+            },
+        ],
+        handoffCommands: [],
+    };
+}
 function visionProtocol() {
     return {
         depth: "deep",
-        interactionMode: "conversation-first-sustained-grill",
+        interactionMode: "delayed-compile-product-interrogation",
         requiredContext: [
             ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
             ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
@@ -74,33 +483,70 @@ function visionProtocol() {
         auditCheckpoints: {
             before: [
                 "Confirm workflow and context indexes exist.",
-                "Start in conversation mode and ask the next useful vision question before writing artifacts.",
+                "Load CURRENT_STATE.yaml when present to avoid stale stage routing.",
+                "Enter interview mode as product partner, requirements interrogator, and intent compiler.",
+                "Ask the next useful vision question before writing artifacts; do not start by creating or updating durable vision files.",
+                "If resuming, load only current vision/context state needed to continue the conversation and identify any stale current_question.",
             ],
             during: [
                 "Ask one focused question at a time and make each question depend on the previous answer.",
-                "Cover mandatory vision dimensions before validation handoff.",
+                "Keep brainstorming in conversation memory until a meaningful checkpoint, pause request, or compile readiness boundary.",
+                "Challenge thin, generic, contradictory, or implementation-shaped answers until the strategic product intent is clear.",
+                "Cover mandatory vision and proto-readiness dimensions before validation handoff.",
                 "Provide concrete examples or options when the user is stuck.",
             ],
             after: [
-                "Persist artifacts only after stable answers, explicit save request, or readiness checkpoint.",
-                "Handoff to validation only after mandatory coverage, unresolved blockers are named, and the user confirms readiness.",
+                "Persist artifacts only in checkpoint mode or compile mode after stable answers, explicit save request, or readiness checkpoint.",
+                "Compile durable VISION artifacts only after proto_readiness.status can be ready or blockers are explicit and the user confirms the interview can stop.",
+                "Handoff to validation only after mandatory coverage, proto-readiness, unresolved blockers, and user readiness are explicit.",
+                "When handing off, mark the vision session active or reviewed, clear stale current_question when answered, and update CURRENT_STATE.yaml.",
                 "Confirm no validation, prototype, spec, change, or runtime artifacts were created.",
             ],
         },
         antiPatterns: [
+            "Do not write durable vision artifacts after every user answer.",
             "Do not open by writing vision artifacts before the conversation has stable answers.",
             "Do not create validation rankings during vision work.",
+            "Do not create prototype prompt packs during vision work; record proto-readiness inputs and hand off to /ow:proto later.",
             "Do not create specs, changes, tasks, or teams from a vision session.",
             "Do not batch many interview questions into one turn.",
             "Do not hand off to validation after a fixed small number of questions.",
+            "Do not hide thin or conflicted answers as polished product truth.",
         ],
         internalSections: [
+            {
+                tag: "vision_role",
+                items: [
+                    "Act as product partner: improve the product thesis, surface sharper alternatives, and protect the user from weak strategic defaults.",
+                    "Act as requirements interrogator: probe unclear assumptions, contradictions, target user ambiguity, trust boundaries, and success criteria.",
+                    "Act as intent compiler: convert conversation into structured strategic_core, product_system_seed, proto_readiness, and coverage fields only when the intent is stable.",
+                ],
+            },
+            {
+                tag: "interaction_modes",
+                items: [
+                    "Interview mode is the default: ask one focused question at a time and do not write durable .openworkflow/vision artifacts after each answer.",
+                    "Checkpoint mode writes a lightweight durable snapshot only when the user asks to save, the session is pausing, a topic has closed, or a load-bearing ambiguity must not be lost.",
+                    "Compile mode writes VISION.md, VISION_CONTRACT.yaml, VISION_SESSION.yaml, and NOTE.md only after mandatory discovery coverage and proto-readiness are sufficient or explicitly blocked.",
+                ],
+            },
+            {
+                tag: "agent_first_consumer",
+                items: [
+                    "Treat the next implementing Agent as the first consumer of vision artifacts.",
+                    "Before persistence or handoff, make the compact vision state answer: current state, read-first pointers, source-of-truth artifact, unresolved blockers, safe write boundary, proto-readiness, validation target, and next command.",
+                    "The vision_delta must preserve enough handoff intelligence for a low-context Agent: one sentence, users, core problem, goals, non-goals, quality bar, AI-native role, success signals, and failure signals.",
+                    "The strategic_core and product_system_seed must let /ow:proto generate prototype directions without inventing the core strategy.",
+                    "If those handoff fields are thin, continue the conversation or record explicit unresolved questions instead of presenting the artifact as ready.",
+                ],
+            },
             {
                 tag: "conversation_first",
                 items: [
                     "Treat /ow:vision as a focused product conversation, not an artifact fill-out task.",
                     "Ask exactly one question unless the user explicitly requests a summary or save checkpoint.",
                     "Let each answer drive the next deeper question; do not run a generic questionnaire mechanically.",
+                    "A long interview is acceptable when it improves product truth; auditability is preserved through checkpoints and compile, not per-answer file churn.",
                 ],
             },
             {
@@ -114,199 +560,700 @@ function visionProtocol() {
                     "Cover privacy, data, sharing, and retention assumptions.",
                     "Cover alternatives or competing mental models.",
                     "Cover success signals and failure signals.",
+                    "Cover prototype direction seeds and prompt constraints needed by /ow:proto.",
+                ],
+            },
+            {
+                tag: "proto_readiness_gate",
+                items: [
+                    "VISION.md is ready only when /ow:proto can derive 3-5 strategically distinct prototype directions without inventing the product strategy.",
+                    "Before compile, verify target user, behavior change, mechanism, differentiator, boundary conditions, trust controls, anti-goals, strongest success signal, failure signals, prototype direction seeds, prompt constraints, and validation target.",
+                    "If proto_readiness.status is missing or thin, keep interviewing or record explicit blockers; do not hand off as ready.",
+                    "If the user asks for prototype prompts directly and proto-readiness is ready, hand off to /ow:proto; otherwise continue the vision interview.",
+                ],
+            },
+            {
+                tag: "readiness_gate",
+                items: [
+                    "Do not hand off to /ow:validation until mandatory coverage is addressed, proto-readiness is ready or explicitly blocked, unresolved questions are explicit, and the user confirms readiness.",
+                    "If a dimension is thin, ask another targeted question instead of writing a final artifact.",
+                    "Vision readiness is based on strategic depth, proto-readiness, and user confirmation, not on a fixed number of turns.",
+                ],
+            },
+            {
+                tag: "artifact_checkpoint",
+                items: [
+                    "Write VISION_SESSION.yaml, VISION_CONTRACT.yaml, VISION.md, or context updates only after stable answers, explicit checkpoint request, pause boundary, or compile readiness.",
+                    "Keep brainstorming and tentative hypotheses in NOTE.md or unresolved_questions rather than presenting them as stable product truth.",
+                    "Summarize at meaningful checkpoints before persisting durable vision state.",
+                ],
+            },
+        ],
+        handoffCommands: ["/ow:validation"],
+    };
+}
+function validationProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "prototype-validation-target-compiler",
+        requiredContext: [
+            ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
+            ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
+            ".openworkflow/vision/VISION_CONTRACT.yaml",
+        ],
+        optionalContext: [
+            ".openworkflow/validation/VALIDATION_INDEX.yaml",
+            ".openworkflow/vision/VISION.md",
+            ".openworkflow/context/CONTEXT.md",
+            ".openworkflow/context/CONTEXT_MAP.yaml",
+        ],
+        forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**"],
+        allowedOutputs: [
+            ".openworkflow/validation/VALIDATION_INDEX.yaml",
+            ".openworkflow/validation/<id>/VALIDATION.yaml",
+            ".openworkflow/validation/<id>/NOTE.md",
+        ],
+        forbiddenOutputs: [
+            ".openworkflow/prototypes/**",
+            ".openworkflow/specs/**",
+            ".openworkflow/changes/**",
+            ".openworkflow/runtime/**",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Confirm a vision contract exists and is proto-ready enough to validate.",
+                "Load CURRENT_STATE.yaml when present.",
+                "Load only vision, validation index, and build-validation context.",
+                "Return to /ow:vision when missing vision fields would force /ow:proto to invent product strategy.",
+            ],
+            during: [
+                "Select exactly one central uncertainty for the next prototype to reduce.",
+                "Define target_behavior and the minimum prototype_experiment needed to observe it.",
+                "Write observable_signals for pass, fail, and ambiguous evidence.",
+                "Write decision_rules for continue, revise, pivot, stop, and needs_more_evidence.",
+                "Record vision_gaps and agent_readiness_gate without generating prototype artifacts.",
+            ],
+            after: [
+                "Record central_uncertainty, hypothesis, target_behavior, prototype_experiment, observable_signals, decision_rules, vision_gaps, and agent_readiness_gate.",
+                "Set agent_readiness_gate.status to ready_for_proto, thin_validation, stale_validation, or return_to_vision.",
+                "Update CURRENT_STATE.yaml with current_validation, active_stage validation, and the next command.",
+                "Mark superseded validation targets accordingly when a new validation target replaces them.",
+                "Confirm no prototype, spec, change, or runtime artifacts were created.",
+            ],
+        },
+        antiPatterns: [
+            "Do not treat feature classification or backlog ranking as the validation outcome.",
+            "Do not generate prototype prompts, images, HTML, specs, changes, or runtime artifacts.",
+            "Do not hide missing vision evidence by writing a polished but unsupported validation target.",
+            "Do not select multiple unrelated validation targets in one artifact.",
+        ],
+        handoffCommands: ["/ow:proto", "/ow:vision"],
+    };
+}
+function prototypeProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "image-first-strategic-proto-prompt-pack",
+        requiredContext: [
+            ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
+            ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
+        ],
+        optionalContext: [
+            ".openworkflow/vision/VISION_CONTRACT.yaml",
+            ".openworkflow/vision/VISION.md",
+            ".openworkflow/validation/VALIDATION_INDEX.yaml",
+            ".openworkflow/validation/**/VALIDATION.yaml",
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            ".openworkflow/context/CONTEXT.md",
+            ".openworkflow/context/CONTEXT_MAP.yaml",
+            "package.json",
+        ],
+        forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**", ".openworkflow/specs/**"],
+        allowedOutputs: [
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            ".openworkflow/prototypes/<id>/PROTO_PROMPT_PACK.yaml",
+            ".openworkflow/prototypes/<id>/PROTO_PROMPT_PACK.md",
+            ".openworkflow/prototypes/<id>/REVIEW_PLAN.md",
+            ".openworkflow/prototypes/<id>/EVIDENCE.yaml",
+            ".openworkflow/prototypes/<id>/NOTE.md",
+            ".openworkflow/prototypes/<id>/images/**",
+            ".openworkflow/decisions/DECISION_INDEX.yaml",
+            ".openworkflow/decisions/<id>/DECISION.yaml",
+            ".openworkflow/decisions/<id>/NOTE.md",
+        ],
+        conditionalOutputs: [
+            ".openworkflow/validation/VALIDATION_INDEX.yaml",
+            ".openworkflow/validation/<id>/VALIDATION.yaml",
+            ".openworkflow/validation/<id>/NOTE.md",
+        ],
+        forbiddenOutputs: [
+            ".openworkflow/prototypes/<id>/review.html",
+            ".openworkflow/specs/**",
+            ".openworkflow/changes/**",
+            ".openworkflow/runtime/**",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Act as the user-facing orchestrator for the internal proto pipeline: proto-preflight, /ow:build-proto-prompt or compatible /ow:vision2prompt compiler path, then /ow:prompt2proto/build-prototype consumption.",
+                "Load vision and current validation context; validation is required before prototype generation.",
+                "If current_validation is missing, auto-run /ow:validation first and write VALIDATION.yaml, NOTE.md, and VALIDATION_INDEX.yaml with trigger.mode agent_auto, requested_command /ow:proto, and reason missing_current_validation.",
+                "Proceed only after validation_input.mode can reference a durable validation artifact; do not use ephemeral vision_only validation context.",
+                "Verify vision and validation artifact quality before prompt work; if either is missing, thin, stale, or not strong enough for high-quality prototype prompts, route back to /ow:vision with focused follow-up questions instead of generating prompts.",
+                "If the user has not specified the number of strategically different prototype directions, askUserQuestion for the count; use 3 only when the user explicitly delegates that decision to the agent.",
+                "Extract the strategic core: target user, behavior change, mechanism, differentiator, boundary conditions, and central uncertainty.",
+            ],
+            during: [
+                "Before strategic directions, require /ow:build-proto-prompt or compatible /ow:vision2prompt compiler path to infer product_experience_model: product archetype, primary canvas, information architecture, domain objects, task loop, interaction states, data realism, visual language, and anti-generic constraints.",
+                "Before /ow:prompt2proto, require prototype_system_contract so stable app shell, navigation, data vocabulary, object anatomy, action bar, audit pattern, copy tone, and allowed deltas are explicit.",
+                "Treat scenarios such as planning, incident, or capacity as possible modules, layers, workflows, or states inside one product shell unless they truly imply different product forms.",
+                "Internally trigger /ow:build-proto-prompt or compatible /ow:vision2prompt compiler path to generate 5-8 strategic prototype hypotheses, select the resolved direction count, and write all multi-direction, multi-image prompt text.",
+                "Do not internally trigger /ow:prompt2proto until prompt_text_manifest.status is ready_for_image_generation and every selected direction has concrete screen prompts.",
+                "Do not internally trigger /ow:prompt2proto until prototype_system_contract exists, prompt_pack_integrity_gate.status and prototype_reality_gate.status are pass and quality_rubric.prompt_executability is pass.",
+                "Do not internally trigger /ow:prompt2proto until post_validate.status is pass for resolved_count 2 or more, or skipped when the user explicitly requested exactly one strategic direction.",
+                "If post_validate.status is fail, or if prompt_pack_integrity_gate, prototype_reality_gate, prompt executability, prototype_system_contract, paragraph quality, or philosophy readiness fails, route back through /ow:build-proto-prompt or compatible /ow:vision2prompt prompt repair instead of starting image generation.",
+                "After prompt_text_manifest.status is ready_for_image_generation and post_validate.status is pass or skipped, Batch-generate prototype images by internally triggering /ow:prompt2proto/build-prototype only after ready prompt-pack artifacts exist; build-prototype consumes the ready pack and must not recompile vision into prompt text.",
+                "Recommend the first direction to generate based on risk reduction, observability, feasibility, and closeness to the success signal.",
+            ],
+            after: [
+                "Write PROTO_PROMPT_PACK.yaml, PROTO_PROMPT_PACK.md, REVIEW_PLAN.md, compact EVIDENCE.yaml, prompt_text_manifest, and image_generation collection state.",
+                "Record review evidence and a decision audit record internally after prompt-pack evidence changes.",
+                "Refresh prototype SUMMARY.yaml when summary_policy is configured and update CURRENT_STATE.yaml with current_prototype, last_decision, and next_command.",
+                "Confirm no HTML, design, spec, change, team, persistence, or production hardening was created.",
+            ],
+        },
+        antiPatterns: [
+            "Do not generate HTML, CSS, local runnable apps, or implementation tasks from /ow:proto.",
+            "Do not treat visual style variants as strategic directions.",
+            "Do not hide missing validation or proceed in ephemeral vision_only mode.",
+            "Do not convert prompt packs into production specs or change backlogs.",
+            "Do not create design, specs, changes, or teams from unaccepted prompt-pack evidence.",
+            "Do not ask the user to manually invoke /ow:decision after prototype work; record the decision audit internally.",
+        ],
+        internalSections: [
+            {
+                tag: "internal_proto_pipeline",
+                items: [
+                    "/ow:proto is the only user-facing command in this chain; /ow:build-proto-prompt, /ow:vision2prompt compatibility, and /ow:prompt2proto are internal commands.",
+                    "/ow:vision2prompt and /ow:prompt2proto are internal commands; do not expose them as normal user-facing handoffs from /ow:proto.",
+                    "Run proto-preflight first, then /ow:build-proto-prompt or compatible /ow:vision2prompt compiler path, then /ow:prompt2proto/build-prototype consumption.",
+                    "Record internal_pipeline.stages with stage ids proto-preflight, build-proto-prompt or vision2prompt-compatible compiler, and prompt2proto in EVIDENCE.yaml.",
+                    "Do not expose /ow:build-proto-prompt, /ow:vision2prompt, or /ow:prompt2proto as normal user-facing handoffs.",
+                ],
+            },
+            {
+                tag: "validation_consumption",
+                items: [
+                    "If validation artifacts are absent but a vision exists, auto-run /ow:validation first and persist VALIDATION.yaml, NOTE.md, and VALIDATION_INDEX.yaml.",
+                    "Auto validation must set trigger.mode: agent_auto, trigger.requested_command: /ow:proto, and trigger.reason: missing_current_validation.",
+                    "If VALIDATION.yaml exists, consume it and preserve central_uncertainty, prototype_experiment, observable_signals, decision_rules, and include/exclude boundaries.",
+                    "Before prompt generation, verify that both VISION.md or VISION_CONTRACT.yaml and VALIDATION.yaml are high-quality enough to produce high-quality prototype prompts.",
+                    "If source quality is insufficient, record preflight_quality_gate.can_proceed false and hand back to /ow:vision with targeted missing questions.",
+                    "If validation conflicts with vision, stop for a decision instead of broadening scope silently.",
+                ],
+            },
+            {
+                tag: "preflight_quality_gate",
+                items: [
+                    "Set preflight_quality_gate.vision_status and validation_status to missing, thin, or ready before prompt work.",
+                    "Set preflight_quality_gate.can_proceed true only when vision and validation can support high-quality prototype prompt generation without agent invention.",
+                    "When can_proceed is false, set next_command_when_blocked: /ow:vision and include required_followup_questions for the user's supplemental interview.",
+                ],
+            },
+            {
+                tag: "direction_count_policy",
+                items: [
+                    "If the user did not specify NUMBER_OF_TYPES or a strategic direction count, askUserQuestion before generating prompt directions.",
+                    "If the user answers with a count, set source: user_input and resolved_count to that number.",
+                    "If the user delegates direction count to the agent, set source: agent_default_after_user_delegation and resolved_count: 3.",
+                    "Do not silently default to 3 before the user either provides a count or delegates the choice.",
+                ],
+            },
+            {
+                tag: "strategic_prompt_pack",
+                items: [
+                    "Write prompt_pack_type: strategic_proto_prompt_pack.",
+                    "Normalize product domain, primary user, usage context, current alternative, core pain, desired behavior change, strongest success signal, core differentiator, emotional value, functional value, trust requirements, privacy requirements, non-goals, future opportunities, and validation target.",
+                    "Represent strategic_core as target user plus behavior change plus mechanism plus differentiator plus boundary conditions.",
+                    "Write product_experience_model before directions: product_archetype, primary_canvas, information_architecture, domain_object_model, primary_task_loop, interaction_state_model, data_realism_requirements, visual_language, anti_generic_constraints, and category_quality_bar.",
+                    "Use product_experience_model to decide whether source concepts are separate strategic directions or modules, scenarios, layers, states, or workflows inside one product shell.",
+                    "Convert non-goals and category anti-patterns into negative_constraints; explicitly block generic AI dashboards, consulting-report layouts, and card walls when the target product category implies a richer product shell.",
+                    "Generate more candidate hypotheses than needed, then select the resolved direction count with maximum strategic diversity.",
+                    "Each direction must include direction_id, name, strategic_hypothesis, validates, main_risk, distinctness_rationale, prototype_prompt, screen_prompts, and pm_judgment.",
+                ],
+            },
+            {
+                tag: "prompt_text_manifest",
+                items: [
+                    "Write complete prompt text for every selected direction before invoking image generation.",
+                    "Each direction should include multi-image screen prompt text with prompt_id, screen_name, image_role, prompt, and acceptance_criteria.",
+                    "Set prompt_text_manifest.status: ready_for_image_generation only after every selected direction has concrete, directly executable prompt text and screen-bound prompt refs.",
+                    "Do not set prompt_text_manifest.status: ready_for_image_generation when prompt_pack_integrity_gate, prototype_reality_gate, screen_manifest coverage, or quality_rubric.prompt_executability is missing or failing.",
+                ],
+            },
+            {
+                tag: "prompt_pack_handoff_gate",
+                items: [
+                    "Before /ow:prompt2proto, require prompt_pack_integrity_gate.status: pass.",
+                    "Before /ow:prompt2proto, require prototype_reality_gate.status: pass.",
+                    "Before /ow:prompt2proto, require screen_manifest entries and direction screen_prompts to resolve by target_screen_id.",
+                    "Before /ow:prompt2proto, require quality_rubric.prompt_executability.status: pass.",
+                    "If any required gate fails or is missing, keep image_generation.status: not_started and repair inside /ow:vision2prompt.",
+                ],
+            },
+            {
+                tag: "post_validate_gate",
+                items: [
+                    "Run prompt asset post-validation after prompt_text_manifest.status is ready_for_image_generation and before /ow:prompt2proto.",
+                    "Require post_validate.status: pass when direction_count_policy.resolved_count is 2 or more.",
+                    "Set post_validate.status: skipped only when direction_count_policy.resolved_count is 1 because the user explicitly requested exactly one strategic direction.",
+                    "Do not start image_generation or invoke /ow:prompt2proto when post_validate.status is fail; route back to /ow:vision2prompt prompt repair.",
+                ],
+            },
+            {
+                tag: "image_generation",
+                items: [
+                    "After prompt_text_manifest.status is ready_for_image_generation and post_validate.status is pass or skipped, Batch-generate prototype images from the prepared prompt text.",
+                    "Generate image groups by direction and screen prompt; keep each generated image linked to direction_id and prompt_id.",
+                    "Record image_generation.status, batch_strategy, generated_images, and collection_notes in EVIDENCE.yaml.",
+                    "Do not use image generation as a substitute for missing strategy or incomplete prompt text.",
+                ],
+            },
+            {
+                tag: "image_only_boundary",
+                items: [
+                    "/ow:proto creates prompt packs and image groups for high-fidelity static prototype images.",
+                    "Do not write HTML, CSS, runnable prototypes, production code, deployment config, auth, persistence, or team runtime.",
+                    "Hand off to /ow:tune when generated images or accepted baseline screens need refinement.",
+                ],
+            },
+            {
+                tag: "review_evidence",
+                items: [
+                    "Record selected direction, user feedback, accepted elements, rejected elements, tune requests, and recommendation.",
+                    "Use recommendation continue, tune, pivot, stop, or needs_more_evidence.",
+                    "Reference generated images by path when present; do not embed large binary evidence.",
+                ],
+            },
+            {
+                tag: "internal_decision_audit",
+                items: [
+                    "After creating or revising prototype evidence, write or update a decision audit record without asking the user to invoke /ow:decision.",
+                    "Use revise when the user asks for another tuning pass, continue when the user explicitly accepts evidence for design, pivot or stop when explicitly directed, and needs_more_evidence when evidence is inconclusive.",
+                    "Keep decision audit output in .openworkflow/decisions/** and do not expose internal bookkeeping as the user-facing workflow step.",
+                ],
+            },
+        ],
+        handoffCommands: ["/ow:tune", "/ow:design", "/ow:validation"],
+    };
+}
+function vision2PromptProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "internal-vision-to-strategic-prompt-text",
+        requiredContext: [
+            ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
+            ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
+            ".openworkflow/vision/VISION_CONTRACT.yaml",
+            ".openworkflow/vision/VISION.md",
+            ".openworkflow/validation/**/VALIDATION.yaml",
+        ],
+        optionalContext: [
+            ".openworkflow/context/CONTEXT.md",
+            ".openworkflow/context/CONTEXT_MAP.yaml",
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+        ],
+        forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**", ".openworkflow/specs/**"],
+        allowedOutputs: [
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            ".openworkflow/prototypes/<id>/PROTO_PROMPT_PACK.yaml",
+            ".openworkflow/prototypes/<id>/PROTO_PROMPT_PACK.md",
+            ".openworkflow/prototypes/<id>/REVIEW_PLAN.md",
+            ".openworkflow/prototypes/<id>/EVIDENCE.yaml",
+            ".openworkflow/prototypes/<id>/NOTE.md",
+        ],
+        forbiddenOutputs: [
+            ".openworkflow/prototypes/<id>/images/**",
+            ".openworkflow/prototypes/<id>/review.html",
+            ".openworkflow/specs/**",
+            ".openworkflow/changes/**",
+            ".openworkflow/runtime/**",
+        ],
+        auditCheckpoints: {
+            before: [
+                "Run only as an internal stage after /ow:proto preflight has confirmed vision and validation are ready.",
+                "Consume durable VISION and VALIDATION artifacts; do not ask broad product questions or generate images.",
+                "Adopt the co-founder plus senior product-manager perspective engine before running references; the references are tools, not a checklist that proves quality by itself.",
+                "Load skills/build-prototype/references/strategic-prompt-pack-protocol.md, then run references/vision2prompt/01_input_contract.md through 07_quality_rubric.md in order before writing PROTO_PROMPT_PACK.",
+                "Resolve direction_count_policy before writing prompt text; use resolved_count from /ow:proto preflight.",
+                "Prepare to infer product_experience_model before strategic directions; do not start directions from scenario labels or visual style variants.",
+                "Prepare to write prototype_system_contract before screen prompt text so stable app shell, navigation taxonomy, data vocabulary, object anatomy, action bar, audit pattern, copy tone, and allowed screen deltas are explicit.",
+                "Prepare to write dailin-grade long-form prompt paragraphs; direction and screen prompt text must include journey, interaction behavior, system response, trust controls, anti-goals, visual direction, desired user feeling, and concrete content.",
+                "Prepare to run prompt_pack_integrity_gate, prototype_reality_gate, quality_rubric.prompt_executability, and post_validate after prompt assets are complete; do not hand off to /ow:prompt2proto while any required gate is missing or failing.",
+            ],
+            during: [
+                "Apply the OW vision2prompt reference pipeline inside OW artifacts and keep intermediate reasoning outputs compactly represented in PROTO_PROMPT_PACK.yaml or NOTE.md.",
+                "Record the intermediate pipeline outputs: perspective engine, normalized input contract, vision decomposition, candidate strategic hypotheses, product experience model, prototype_system_contract, screen_manifest, prototype prompt schema, output manifest, and quality rubric.",
+                "Each selected direction must carry product_thesis, user_transformation, reason_to_exist, differentiated product form, and pm_judgment before screen prompt anatomy.",
+                "Infer product_experience_model from VISION and VALIDATION: product archetype, primary canvas, information architecture, domain object model, primary task loop, interaction state model, data realism requirements, visual language, anti-generic constraints, and category quality bar.",
+                "Decide whether scenario names are separate strategic product forms or modules, layers, workflows, or states inside one product shell; do not split directions by scenario labels alone.",
+                "Generate more candidate hypotheses than needed and select the resolved direction count for maximum strategic diversity.",
+                "Only select directions that differ by product form, product loop, initiation trigger, interaction model, emotional driver, retention mechanism, validation metric, or main risk.",
+                "Write prototype_system_contract cross-screen invariants before complete multi-image prompt text so downstream prompt2proto preserves technical screen coherence instead of inventing it from visual judgment.",
+                "Write complete multi-image prompt text for every selected direction with screen_prompts, negative_prompt, example_copy, and acceptance criteria tied to screen_manifest target_screen_id values.",
+                "Populate quality_rubric.prompt_paragraph_quality and prompt_text_manifest.paragraph_quality_status before handoff; fail closed when paragraph_quality_dimensions are missing.",
+                "Do not set prompt_text_manifest.status to ready_for_image_generation until prompt_pack_integrity_gate.status, prototype_reality_gate.status, quality_rubric.prompt_executability.status, and prompt_text_manifest.paragraph_quality_status are pass.",
+                "After prompt text is ready, run the deterministic post_validate gate over strategic_fingerprint dimensions when resolved_count is 2 or more.",
+                "When the user explicitly requested exactly one strategic direction, set post_validate.status: skipped and record the skip reason instead of running diversity comparison.",
+            ],
+            after: [
+                "Write PROTO_PROMPT_PACK.yaml, PROTO_PROMPT_PACK.md, REVIEW_PLAN.md, and EVIDENCE.yaml with product_experience_model, prototype_system_contract, screen_manifest, screen_prompts, prompt_pack_integrity_gate, prototype_reality_gate, quality_rubric, prompt_text_manifest.status ready_for_image_generation, and post_validate status.",
+                "Hand internally to /ow:prompt2proto only when prototype_system_contract is present, prompt_pack_integrity_gate.status and prototype_reality_gate.status are pass, quality_rubric.prompt_executability.status is pass, prompt_text_manifest.paragraph_quality_status is pass, and post_validate.status is pass or skipped.",
+                "If any readiness gate fails, keep handoff blocked and repair prompt directions through /ow:vision2prompt.",
+                "Record internal_pipeline stage vision2prompt status and outputs.",
+                "Do not generate images; hand internally to /ow:prompt2proto.",
+            ],
+        },
+        antiPatterns: [
+            "Do not expose /ow:vision2prompt as a user-facing workflow step.",
+            "Do not generate prototype images from this stage.",
+            "Do not invent strategy when vision or validation is thin; return control to /ow:proto preflight.",
+            "Do not treat dailin-derived references or YAML field completion as quality when the prompt lacks product thesis, user transformation, reason-to-exist, or design philosophy.",
+            "Do not treat screen_prompts[].prompt as image-generation-ready when it reads like a screen-state instruction instead of a dailin-grade prototype-generation brief.",
+            "Do not hand off to /ow:prompt2proto when post_validate.status is fail or missing for multi-direction prompt packs.",
+            "Do not hand off to /ow:prompt2proto when prototype_system_contract, prompt_pack_integrity_gate, prototype_reality_gate, screen_manifest coverage, or prompt executability is missing or failing.",
+            "Do not create HTML, specs, changes, or runtime artifacts.",
+        ],
+        internalSections: [
+            {
+                tag: "internal_command_boundary",
+                items: [
+                    "/ow:vision2prompt is internal and is invoked by /ow:proto, not by the user.",
+                    "Its only job is strategic prompt text generation from ready vision and validation artifacts.",
+                    "Its output must be ready for /ow:prompt2proto consumption.",
+                ],
+            },
+            {
+                tag: "perspective_engine",
+                items: [
+                    "Start from a co-founder plus 15-year senior product-manager perspective before executing the reference pipeline.",
+                    "Treat dailin-derived references as tools for judgment, not a checklist whose completion alone proves quality.",
+                    "Ask what product should exist, why this prototype matters, which form best expresses the vision, and what user transformation should become visible.",
+                    "Every selected direction must include product_thesis, user_transformation, differentiated product form, reason_to_exist, and pm_judgment.",
+                    "Reject complete-but-soulless prompt paragraphs that list screens but do not make a product argument.",
+                ],
+            },
+            {
+                tag: "vision2prompt_reference_pipeline",
+                items: [
+                    "Run skills/build-prototype/references/strategic-prompt-pack-protocol.md as the governing protocol.",
+                    "Load references/vision2prompt/01_input_contract.md to normalize vision, validation, direction count, target tool, fidelity, constraints, and missing inputs.",
+                    "Load references/vision2prompt/02_vision_decomposition.md before strategic hypotheses.",
+                    "Load references/vision2prompt/03_strategy_hypothesis_generation.md to create 5-8 candidates and select materially distinct directions.",
+                    "Load references/vision2prompt/04_product_system_extraction.md before writing screen_manifest or product_experience_model.",
+                    "Load references/vision2prompt/05_prototype_prompt_schema.md before writing direction screen_prompts.",
+                    "Load references/vision2prompt/06_output_templates.md so YAML remains source of truth and Markdown remains a readable view.",
+                    "Load references/vision2prompt/07_quality_rubric.md before marking prompt_text_manifest ready.",
+                ],
+            },
+            {
+                tag: "prompt_paragraph_quality",
+                items: [
+                    "Before writing prompt text, run the mapped OW equivalents of dailin 01-06 references in order; references are mandatory generation tools, not optional background.",
+                    "Every direction prototype_prompt must be a dailin-grade long-form prototype-generation brief, not a short image prompt.",
+                    "Every screen_prompts[].prompt must include journey stage, interaction behavior, system response, trust controls, anti-goals, visual direction, desired user feeling, concrete content, and screen-bound acceptance criteria.",
+                    "quality_rubric.prompt_paragraph_quality must record pass/fail evidence for product context, target user, journey, screens/components, interaction/system response, concrete content, trust/user control, visual direction, anti-goals, desired user feeling, and the perspective engine.",
+                    "prompt_text_manifest.paragraph_quality_status must be pass before /ow:prompt2proto handoff, with paragraph_quality_dimensions listing the dimensions that were checked.",
+                ],
+            },
+            {
+                tag: "product_experience_model",
+                items: [
+                    "Before directions, write product_experience_model with product_archetype, primary_canvas, information_architecture, domain_object_model, primary_task_loop, interaction_state_model, data_realism_requirements, visual_language, anti_generic_constraints, and category_quality_bar.",
+                    "Use the model to preserve target product category reality from VISION-only input.",
+                    "Do not treat modules, scenarios, layers, workflows, or interaction states as separate strategic directions unless they imply different product forms or loops.",
+                    "Block generic AI dashboard, report-screen, or card-wall drift through anti_generic_constraints and negative_constraints.",
+                ],
+            },
+            {
+                tag: "prototype_system_contract",
+                items: [
+                    "Before screen prompts, write prototype_system_contract as the technical multi-screen coherence contract.",
+                    "The contract must cover stable_app_shell, navigation_taxonomy, data_vocabulary, domain_object_anatomy, object_detail_anatomy, action_bar_contract, audit_trust_pattern, copy_tone, and allowed_screen_deltas.",
+                    "Use the contract to separate cross-screen invariants from screen-specific workflow deltas.",
+                    "Treat multi-screen drift as a prompt-pack consistency issue, not as provider image quality, human visual review, or visual parity proof.",
                 ],
             },
             {
-                tag: "readiness_gate",
+                tag: "prompt_pack_readiness_gate",
                 items: [
-                    "Do not hand off to /ow:validation until mandatory coverage is addressed, unresolved questions are explicit, and the user confirms readiness.",
-                    "If a dimension is thin, ask another targeted question instead of writing a final artifact.",
-                    "Vision readiness is based on coverage and user confirmation, not on a fixed number of turns.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prompt_pack_integrity_gate.status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prototype_reality_gate.status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prototype_system_contract records cross-screen invariants and allowed deltas.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until quality_rubric.prompt_executability.status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prompt_text_manifest.paragraph_quality_status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until every direction screen_prompt target_screen_id resolves to screen_manifest.",
+                    "When a gate fails or is missing, set image_generation.status: not_started and repair through /ow:vision2prompt before handoff.",
                 ],
             },
             {
-                tag: "artifact_checkpoint",
+                tag: "post_validate_gate",
                 items: [
-                    "Write VISION_SESSION.yaml, VISION_CONTRACT.yaml, VISION.md, or context updates only after stable answers or an explicit checkpoint request.",
-                    "Keep brainstorming and tentative hypotheses in NOTE.md or unresolved_questions rather than presenting them as stable product truth.",
-                    "Summarize at meaningful checkpoints before persisting durable vision state.",
+                    "Run post_validate after prompt assets are ready and before /ow:prompt2proto handoff.",
+                    "For resolved_count 2 or more, require post_validate.status: pass before image generation.",
+                    "For resolved_count 1 from explicit user input, set post_validate.status: skipped and record why the diversity gate did not run.",
+                    "For post_validate.status: fail, repair strategic prompt directions inside /ow:vision2prompt instead of invoking /ow:prompt2proto.",
                 ],
             },
         ],
-        handoffCommands: ["/ow:validation"],
+        handoffCommands: ["/ow:prompt2proto"],
     };
 }
-function validationProtocol() {
+function buildProtoPromptProtocol() {
     return {
         depth: "deep",
-        interactionMode: "audit-and-rank",
+        interactionMode: "internal-build-proto-prompt-pack-compiler",
         requiredContext: [
             ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
             ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
             ".openworkflow/vision/VISION_CONTRACT.yaml",
+            ".openworkflow/vision/VISION.md",
+            ".openworkflow/validation/**/VALIDATION.yaml",
         ],
         optionalContext: [
-            ".openworkflow/validation/VALIDATION_INDEX.yaml",
-            ".openworkflow/vision/VISION.md",
             ".openworkflow/context/CONTEXT.md",
             ".openworkflow/context/CONTEXT_MAP.yaml",
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            "skills/build-proto-prompt/SKILL.md",
+            "skills/build-proto-prompt/references/prompt-pack-compiler-protocol.md",
+            "skills/build-proto-prompt/references/output-boundary.md",
         ],
-        forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**"],
+        forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**", ".openworkflow/specs/**"],
         allowedOutputs: [
-            ".openworkflow/validation/VALIDATION_INDEX.yaml",
-            ".openworkflow/validation/<id>/VALIDATION.yaml",
-            ".openworkflow/validation/<id>/NOTE.md",
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            ".openworkflow/prototypes/<id>/PROTO_PROMPT_PACK.yaml",
+            ".openworkflow/prototypes/<id>/PROTO_PROMPT_PACK.md",
+            ".openworkflow/prototypes/<id>/REVIEW_PLAN.md",
+            ".openworkflow/prototypes/<id>/EVIDENCE.yaml",
+            ".openworkflow/prototypes/<id>/NOTE.md",
         ],
         forbiddenOutputs: [
-            ".openworkflow/prototypes/**",
+            ".openworkflow/prototypes/<id>/images/**",
+            ".openworkflow/prototypes/<id>/review.html",
             ".openworkflow/specs/**",
             ".openworkflow/changes/**",
             ".openworkflow/runtime/**",
         ],
         auditCheckpoints: {
-            before: ["Confirm a vision contract exists.", "Load only vision and validation index context."],
-            during: ["Classify features as existential, supporting, later, or out of scope.", "Name the single highest-risk validation question."],
-            after: ["Record the validation target and prototype brief.", "Confirm no prototype, spec, change, or runtime artifacts were created."],
+            before: [
+                "Run only as an internal prompt-pack compiler stage after /ow:proto preflight has confirmed vision and validation are ready.",
+                "Load skills/build-proto-prompt/SKILL.md and its references before compiling PROTO_PROMPT_PACK.",
+                "Adopt the Co-Founder plus Chief PM / senior product strategist perspective engine before generating directions.",
+                "Resolve direction_count_policy before writing prompt text; do not silently default the direction count.",
+                "Prepare to infer product_experience_model before strategic directions, including product archetype, primary canvas, information architecture, domain objects, state model, data realism, visual language, anti-generic constraints, and category quality bar.",
+                "Prepare to write prototype_system_contract before screen prompts, covering stable app shell, navigation taxonomy, data vocabulary, object anatomy, action bar, audit pattern, copy tone, and allowed screen deltas.",
+                "Prepare to write dailin-grade long-form prompt paragraphs; prompt text must include journey, interaction behavior, system response, trust controls, anti-goals, visual direction, desired user feeling, and concrete content.",
+            ],
+            during: [
+                "Compile durable vision and validation into strategic_core, prototype_brief, product_experience_model, prototype_system_contract, screen_manifest, global_design_system_prompt, directions, screen_prompts, review_plan, and readiness gates.",
+                "Each selected direction must carry product_thesis, user_transformation, differentiated product form, reason_to_exist, and pm_judgment before screen prompt anatomy.",
+                "Decide whether source concepts are separate strategic product forms or modules, layers, workflows, or states inside one product shell; do not split directions by scenario labels alone.",
+                "Write prototype_system_contract cross-screen invariants before complete multi-image prompt text so downstream prompt2proto preserves technical screen coherence instead of inventing it from visual judgment.",
+                "Write complete multi-image prompt text for every selected direction with screen_prompts, negative_prompt, example_copy, and acceptance criteria tied to screen_manifest target_screen_id values.",
+                "Populate prompt_pack_integrity_gate, prototype_reality_gate, quality_rubric.prompt_executability, quality_rubric.prompt_paragraph_quality, prompt_text_manifest.paragraph_quality_status, and post_validate before handoff.",
+                "Keep image_generation.status: not_started while this compiler stage runs.",
+            ],
+            after: [
+                "Write PROTO_PROMPT_PACK.yaml, PROTO_PROMPT_PACK.md, REVIEW_PLAN.md, and EVIDENCE.yaml with prototype_system_contract and prompt_text_manifest.status ready_for_image_generation only when all readiness gates pass.",
+                "Hand internally to /ow:prompt2proto only when prototype_system_contract is present, prompt_pack_integrity_gate.status and prototype_reality_gate.status are pass, quality_rubric.prompt_executability.status is pass, prompt_text_manifest.paragraph_quality_status is pass, and post_validate.status is pass or skipped.",
+                "If any readiness gate fails, keep handoff blocked and repair prompt directions through /ow:build-proto-prompt.",
+                "Do not generate images; do not consume provider image output or prototype review artifacts.",
+            ],
         },
         antiPatterns: [
-            "Do not turn feature ranking into a production task list.",
-            "Do not prototype before naming the validation question.",
-            "Do not treat supporting features as blockers for existential validation.",
+            "Do not expose /ow:build-proto-prompt as a user-facing workflow step.",
+            "Do not generate prototype images from this stage.",
+            "Do not consume provider image output, human visual review, visual parity scores, proto2html artifacts, specs, changes, or runtime state.",
+            "Do not invent strategy when vision or validation is thin; return control to /ow:proto preflight.",
+            "Do not treat YAML field completion as quality when the prompt lacks product thesis, user transformation, reason-to-exist, or design philosophy.",
+            "Do not hand off to /ow:prompt2proto when post_validate.status is fail or missing for multi-direction prompt packs.",
+            "Do not hand off to /ow:prompt2proto when prototype_system_contract, prompt_pack_integrity_gate, prototype_reality_gate, screen_manifest coverage, paragraph quality, or prompt executability is missing or failing.",
         ],
-        handoffCommands: ["/ow:proto", "/ow:vision"],
+        internalSections: [
+            {
+                tag: "internal_command_boundary",
+                items: [
+                    "/ow:build-proto-prompt is internal and is invoked by /ow:proto orchestration, not by the user.",
+                    "Its only job is prompt-pack compilation from ready vision and validation artifacts.",
+                    "Its output must be ready for prompt2proto consumption.",
+                    "Keep /ow:vision2prompt compatibility until a later migration candidate explicitly removes or aliases it.",
+                ],
+            },
+            {
+                tag: "prompt_pack_compiler_role",
+                items: [
+                    "Start from a Co-Founder plus Chief PM / senior product strategist perspective before executing the prompt-pack compiler references.",
+                    "Ask what product should exist, why this prototype matters, which form best expresses the vision, and what user transformation should become visible.",
+                    "Every selected direction must include product_thesis, user_transformation, differentiated product form, reason_to_exist, and pm_judgment.",
+                    "Reject complete-but-soulless prompt paragraphs that list screens but do not make a product argument.",
+                ],
+            },
+            {
+                tag: "source_skill_reference",
+                items: [
+                    "Load skills/build-proto-prompt/SKILL.md as the source behavior for this command.",
+                    "Load skills/build-proto-prompt/references/prompt-pack-compiler-protocol.md before writing prompt-pack outputs.",
+                    "Load skills/build-proto-prompt/references/output-boundary.md before handoff.",
+                    "The legacy vision2prompt references may be reused as detailed prompt-generation tools until the compiler references are fully split.",
+                ],
+            },
+            {
+                tag: "prototype_system_contract",
+                items: [
+                    "Write prototype_system_contract as the build-proto-prompt technical multi-screen coherence contract.",
+                    "Required fields: stable_app_shell, navigation_taxonomy, data_vocabulary, domain_object_anatomy, object_detail_anatomy, action_bar_contract, audit_trust_pattern, copy_tone, and allowed_screen_deltas.",
+                    "Use it to tell prompt2proto what must remain stable across screens and what may change by workflow state.",
+                    "This is a consistency contract, not density philosophy, provider image quality, visual review, or visual parity scoring.",
+                ],
+            },
+            {
+                tag: "prompt_pack_readiness_gate",
+                items: [
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prompt_pack_integrity_gate.status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prototype_reality_gate.status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prototype_system_contract records cross-screen invariants and allowed deltas.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until quality_rubric.prompt_executability.status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until prompt_text_manifest.paragraph_quality_status is pass.",
+                    "prompt_text_manifest.status cannot become ready_for_image_generation until every direction screen_prompt target_screen_id resolves to screen_manifest.",
+                    "When a gate fails or is missing, set image_generation.status: not_started and repair through /ow:build-proto-prompt before handoff.",
+                ],
+            },
+            {
+                tag: "downstream_boundary",
+                items: [
+                    "Do not generate images.",
+                    "Do not perform human visual review.",
+                    "Do not claim visual reference parity.",
+                    "Do not create proto2html, specs, changes, or runtime artifacts.",
+                    "Do not narrow build-prototype behavior in this command-boundary candidate.",
+                ],
+            },
+        ],
+        handoffCommands: ["/ow:prompt2proto"],
     };
 }
-function prototypeProtocol() {
+function prompt2ProtoProtocol() {
     return {
         depth: "deep",
-        interactionMode: "classified-prototype-creation",
+        interactionMode: "internal-prompt-text-to-prototype-images",
         requiredContext: [
             ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
             ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
-            ".openworkflow/validation/VALIDATION_INDEX.yaml",
+            ".openworkflow/prototypes/**/PROTO_PROMPT_PACK.yaml",
+            ".openworkflow/prototypes/**/EVIDENCE.yaml",
         ],
         optionalContext: [
-            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
             ".openworkflow/validation/**/VALIDATION.yaml",
-            ".openworkflow/vision/VISION_CONTRACT.yaml",
-            "package.json",
+            ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            ".openworkflow/prototypes/**/NOTE.md",
         ],
         forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**", ".openworkflow/specs/**"],
         allowedOutputs: [
             ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
             ".openworkflow/prototypes/<id>/EVIDENCE.yaml",
             ".openworkflow/prototypes/<id>/NOTE.md",
-            ".openworkflow/prototypes/<id>/review.html",
-            ".openworkflow/prototypes/<id>/evidence/**",
+            ".openworkflow/prototypes/<id>/images/**",
             ".openworkflow/decisions/DECISION_INDEX.yaml",
             ".openworkflow/decisions/<id>/DECISION.yaml",
             ".openworkflow/decisions/<id>/NOTE.md",
-            ".openworkflow/decisions/<id>/review.html",
         ],
-        forbiddenOutputs: [".openworkflow/specs/**", ".openworkflow/changes/**", ".openworkflow/runtime/**"],
+        forbiddenOutputs: [
+            ".openworkflow/prototypes/<id>/review.html",
+            ".openworkflow/specs/**",
+            ".openworkflow/changes/**",
+            ".openworkflow/runtime/**",
+        ],
         auditCheckpoints: {
             before: [
-                "Confirm validation target exists.",
-                "Classify prototype mode before implementation: visual, interaction, technical feasibility, 3D/material, workflow, or data/logic.",
-                "Detect reference inputs: image, URL, screenshot, HTML/CSS source, existing artifact, or design-system hint.",
+                "Run only as an internal stage after /ow:vision2prompt has written prompt_text_manifest.status ready_for_image_generation, prompt_text_manifest.paragraph_quality_status pass, and post_validate.status pass or skipped.",
+                "Treat prompt2proto as the internal build-prototype consumption boundary: consume ready PROTO_PROMPT_PACK artifacts and do not recompile vision or validation into prompt text.",
+                "Load prepared prompt text and verify every selected direction has screen_prompts before image generation.",
+                "Verify prototype_system_contract, prompt_pack_integrity_gate.status: pass, prototype_reality_gate.status: pass, quality_rubric.prompt_executability.status: pass, prompt_text_manifest.paragraph_quality_status: pass, and screen_manifest linkage before image generation.",
+                "Apply the build-prototype philosophy engine before image generation: Chief PM plus Principal UI/UX judgment must set product intent, information hierarchy, density calibration, affordance clarity, and UI/UX credibility boundaries.",
+                "Block image generation when prototype_system_contract, prompt_pack_integrity_gate, prototype_reality_gate, prompt executability, paragraph quality, screen_manifest linkage, or post_validate is fail or missing.",
             ],
             during: [
-                "For visual-first prototypes, extract reference patterns and create a high-fidelity static concept with image generation before HTML unless the user explicitly skips it.",
-                "Derive a compact visual direction and token packet before implementation.",
-                "Build only what answers the validation question and keep one command or URL to run the prototype.",
-                "Verify rendered prototypes with browser and screenshot checks before handoff.",
+                "Batch-generate high-fidelity prototype images by direction_id and prompt_id.",
+                "Calibrate prototype instructions so information is visible, grouped, collapsed, delayed, or drilled into based on industry, user role, task risk, screen size, task frequency, and the user's next decision.",
+                "If the prompt pack is missing, thin, stale, incoherent, or lacks prompt2proto philosophy readiness, refuse and route repair back to /ow:build-proto-prompt or compatible /ow:vision2prompt.",
+                "Write one metadata record for every generated image with image_id, direction_id, prompt_id, screen_name, path, source_prompt_ref, generator, and status.",
+                "Do not revise product strategy or prompt text during image generation.",
             ],
             after: [
-                "Record reference analysis, static concept evidence, runnable implementation evidence, verification, self-critique, and known limits separately.",
-                "Write a decision audit record internally after prototype evidence changes.",
-                "Write evidence and result artifacts.",
-                "Confirm no design, spec, change, team, persistence, or production hardening was created.",
+                "Update image_generation.status, generated_images, collection_notes, and internal_pipeline stage prompt2proto outputs.",
+                "Record decision audit internally after image evidence changes.",
+                "Hand the user-facing flow back to /ow:proto for summary and next command.",
             ],
         },
         antiPatterns: [
-            "Do not jump directly to HTML for a visual-first prototype before a visual direction or static concept unless the user explicitly skips it.",
-            "Do not ignore user-provided reference images, URLs, screenshots, or HTML/CSS source.",
-            "Do not force image generation for logic-only, data-flow, API, or technical feasibility prototypes.",
-            "Do not polish the prototype into production code.",
-            "Do not add persistence unless persistence is the validation question.",
-            "Do not create design, specs, changes, or teams from unaccepted prototype work.",
-            "Do not ask the user to manually invoke /ow:decision after prototype work; record the decision audit internally.",
+            "Do not expose /ow:prompt2proto as a user-facing workflow step.",
+            "Do not start image generation when prompt text is not ready.",
+            "Do not start image generation when prompt_text_manifest.paragraph_quality_status is missing, fail, or unchecked.",
+            "Do not start image generation when prompt_pack_integrity_gate or prototype_reality_gate is missing or failing.",
+            "Do not start image generation when prototype_system_contract is missing stable app shell, navigation taxonomy, data vocabulary, object anatomy, action bar, audit trust pattern, copy tone, or allowed screen deltas.",
+            "Do not start image generation when screen prompts are detached from screen_manifest or prompt executability has not passed.",
+            "Do not create HTML, specs, changes, or runtime artifacts.",
+            "Do not allow generated images without per-image metadata.",
         ],
         internalSections: [
             {
-                tag: "prototype_classification",
-                items: [
-                    "Classify the prototype as visual, interaction, technical feasibility, 3D/material, workflow, or data/logic before choosing tools or writing files.",
-                    "Name the validation question, the riskiest assumption, and the smallest success signal.",
-                    "If classification is ambiguous, ask one clarifying question; otherwise proceed with the most likely mode and record the assumption.",
-                ],
-            },
-            {
-                tag: "reference_extraction",
-                items: [
-                    "When the user provides a target image, URL, screenshot, HTML/CSS, or reference artifact, perform reference-pattern extraction before visual generation or HTML implementation.",
-                    "Extract transferable patterns: information architecture, layout rhythm, component grammar, typography posture, palette, motion, interaction details, and anti-patterns to avoid.",
-                    "Record reference analysis as evidence by path or URL; do not paste bulky source or screenshots into YAML.",
-                ],
-            },
-            {
-                tag: "visual_first_path",
-                items: [
-                    "For visual, product-experience, 3D/material, and aesthetic-sensitive interaction prototypes, default to a high-fidelity static concept before runnable HTML.",
-                    "Use image generation as the default first visual pass for composition, mood, material, visual hierarchy, and brand direction unless the user asks to skip image generation.",
-                    "Record visual_concept_policy.image_generation as generated, skipped_by_user, or not_applicable; skipped_by_user requires a concrete skip reason.",
-                    "Discuss or confirm the static concept before spending implementation effort when the user is actively collaborating; if the user asked for autonomous execution, proceed after the concept establishes clear direction.",
-                    "Do not require image generation for data/logic, API, or pure technical feasibility prototypes.",
-                ],
-            },
-            {
-                tag: "design_seed_protocol",
-                items: [
-                    "Do not design from a blank aesthetic when a direction, design system, template seed, or reference exists.",
-                    "Derive a compact visual packet before implementation: background, surface, foreground, muted, border, accent, display font, body font, radius, spacing, motion, and density.",
-                    "Choose domain-appropriate posture: operational tools should be dense and restrained, editorial surfaces can be expressive, games can be playful, and dashboards should avoid marketing hero treatment.",
-                ],
-            },
-            {
-                tag: "implementation_protocol",
+                tag: "build_prototype_consumption_boundary",
                 items: [
-                    "Implement the smallest runnable artifact that validates the current question, not a production app.",
-                    "For HTML prototypes, keep final review surfaces free of designer-only controls unless those controls are part of the validation target.",
-                    "Keep generated assets, screenshots, logs, and review HTML in the prototype evidence folder.",
+                    "build-prototype consumes ready prompt-pack artifacts through prompt2proto.",
+                    "build-prototype must not compile prompt packs from vision or validation.",
+                    "Missing or incoherent prompt packs repair through /ow:build-proto-prompt or compatible /ow:vision2prompt, not inside prompt2proto.",
+                    "Refuse thin, missing, or incoherent prompt packs before image generation.",
                 ],
             },
             {
-                tag: "verification_protocol",
+                tag: "prompt_pack_readiness_gate",
                 items: [
-                    "For rendered HTML or 3D prototypes, run browser verification and capture screenshots or notes for desktop and mobile when practical.",
-                    "Verify that the page is nonblank, core interactions work, primary assets render, text does not overlap, and responsive layout remains coherent.",
-                    "Record known limits separately from observations so downstream decision work can judge evidence quality.",
+                    "Refuse prompt packs whose prompt_pack_integrity_gate.status is not pass.",
+                    "Refuse prompt packs whose prototype_reality_gate.status is not pass.",
+                    "Refuse prompt packs whose quality_rubric.prompt_executability.status is not pass.",
+                    "Refuse prompt packs whose prompt_text_manifest.paragraph_quality_status is not pass.",
+                    "Refuse prompt packs whose prototype_system_contract does not define stable_app_shell, navigation_taxonomy, data_vocabulary, domain_object_anatomy, object_detail_anatomy, action_bar_contract, audit_trust_pattern, copy_tone, and allowed_screen_deltas.",
+                    "Refuse prompt packs whose screen_prompts[].prompt paragraphs omit journey, interaction behavior, system response, trust controls, anti-goals, visual direction, desired user feeling, or concrete content.",
+                    "Refuse prompt packs whose direction screen_prompts do not resolve to screen_manifest target_screen_id values.",
+                    "When a prompt pack is refused, keep image_generation.status: not_started and hand back to /ow:vision2prompt repair.",
                 ],
             },
             {
-                tag: "self_critique",
+                tag: "build_prototype_philosophy_engine",
                 items: [
-                    "Before handoff, critique the prototype across philosophy, hierarchy, execution, specificity, restraint, accessibility, and responsive behavior.",
-                    "Any weak dimension must trigger one repair pass before evidence handoff unless the weakness is intentionally out of scope for the validation question.",
-                    "Record critique findings and repairs as compact evidence references or YAML summary fields.",
+                    "Chief PM plus Principal UI/UX judgment must run before visual translation.",
+                    "The Chief PM protects product intent, user decision context, domain fit, workflow priority, and evidence value.",
+                    "The Principal UI/UX lead protects information hierarchy, density calibration, layout anatomy, affordance clarity, interaction believability, and UI/UX credibility.",
+                    "Density calibration is product and design judgment, not a mechanical prompt dimension or paragraph length target.",
+                    "Use prototype_system_contract for technical screen coherence; use this philosophy engine for visual information judgment.",
+                    "Reject under-specified mockups that hide operational decisions and overstuffed concept posters that make every element equally important.",
                 ],
             },
             {
-                tag: "internal_decision_audit",
+                tag: "image_metadata_contract",
                 items: [
-                    "After creating or revising prototype evidence, write or update a decision audit record without asking the user to invoke /ow:decision.",
-                    "Use revise when the user asks for another tuning pass, continue when the user explicitly accepts evidence for design, pivot or stop when explicitly directed, and needs_more_evidence when evidence is inconclusive.",
-                    "Keep decision audit output in .openworkflow/decisions/** and do not expose internal bookkeeping as the user-facing workflow step.",
+                    "Every generated image must record image_id, direction_id, prompt_id, screen_name, path, and metadata.",
+                    "metadata must include source_prompt_ref, generated_at, generator, generation_status, and review_status.",
+                    "Images without metadata are not valid prototype evidence.",
                 ],
             },
         ],
-        handoffCommands: ["/ow:tune", "/ow:design", "/ow:validation"],
+        handoffCommands: ["/ow:proto"],
     };
 }
 function tuneProtocol() {
     return {
         depth: "deep",
-        interactionMode: "prototype-revision-orchestration",
+        interactionMode: "screen-bound-prototype-refinement",
         requiredContext: [
             ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
             ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
@@ -316,6 +1263,8 @@ function tuneProtocol() {
             ".openworkflow/validation/**/VALIDATION.yaml",
             ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
             ".openworkflow/prototypes/**/EVIDENCE.yaml",
+            ".openworkflow/prototypes/**/PROTO_PROMPT_PACK.yaml",
+            ".openworkflow/prototypes/**/REFINED_PROTO_PROMPT_PACK.yaml",
             ".openworkflow/prototypes/**/NOTE.md",
             ".openworkflow/decisions/DECISION_INDEX.yaml",
             ".openworkflow/decisions/**/DECISION.yaml",
@@ -324,63 +1273,133 @@ function tuneProtocol() {
         forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**", ".openworkflow/specs/**"],
         allowedOutputs: [
             ".openworkflow/prototypes/PROTOTYPE_INDEX.yaml",
+            ".openworkflow/prototypes/<id>/REFINED_PROTO_PROMPT_PACK.yaml",
+            ".openworkflow/prototypes/<id>/REFINED_PROTO_PROMPT_PACK.md",
+            ".openworkflow/prototypes/<id>/REVIEW_PLAN.md",
             ".openworkflow/prototypes/<id>/EVIDENCE.yaml",
             ".openworkflow/prototypes/<id>/NOTE.md",
-            ".openworkflow/prototypes/<id>/review.html",
-            ".openworkflow/prototypes/<id>/evidence/**",
+            ".openworkflow/prototypes/<id>/images/**",
             ".openworkflow/decisions/DECISION_INDEX.yaml",
             ".openworkflow/decisions/<id>/DECISION.yaml",
             ".openworkflow/decisions/<id>/NOTE.md",
-            ".openworkflow/decisions/<id>/review.html",
         ],
-        forbiddenOutputs: [".openworkflow/specs/**", ".openworkflow/changes/**", ".openworkflow/runtime/**"],
+        forbiddenOutputs: [
+            ".openworkflow/prototypes/<id>/review.html",
+            ".openworkflow/specs/**",
+            ".openworkflow/changes/**",
+            ".openworkflow/runtime/**",
+        ],
         auditCheckpoints: {
             before: [
-                "Resolve tune target: /ow:tune and /ow:tune:proto default to the current prototype.",
-                "If no current prototype exists but a current validation target exists, orchestrate prototype creation through /ow:proto behavior.",
-                "Load only the current prototype evidence, relevant validation target, and latest decision audit context.",
+                "Resolve tune target: /ow:tune defaults to the latest approved prototype prompt pack, refined prompt pack, or accepted baseline screen group.",
+                "Require baseline screens, screenshots, screen descriptions, generated images, or an accepted PROTO_PROMPT_PACK plus a tune request.",
+                "Normalize tune inputs before baseline audit: baseline source type, baseline refs, tune request, target form factor, regeneration scope, screen count, locked screens, locked elements, and constraints.",
+                "Record baseline_resolution before auditing: latest approved baseline group id, latest approved baseline ref, lineage, resolution rule, and stale source guard.",
+                "Assign stable source screen ids and target screen ids when the baseline or request does not provide them.",
+                "Load only the baseline prompt pack, current prototype evidence, relevant validation or vision context, and latest decision audit context.",
             ],
             during: [
-                "Apply exactly one focused revision loop from user feedback.",
-                "Preserve M16 prototype evidence separation for concept, implementation, verification, self-critique, and known limits.",
-                "Run required verification for changed rendered artifacts.",
+                "Audit the full baseline screen group before writing refined prompts.",
+                "Carry forward locked screens, locked elements, and accepted improvements from previous tune passes unless the current user request explicitly unlocks them.",
+                "Extract the product system and preserve product thesis, primary loop, component vocabulary, copy tone, AI/system behavior, trust boundaries, and user controls.",
+                "Interpret the tune request as visual refinement, brand alignment, screen-specific correction, feature addition, feature removal, form-factor transformation, group expansion, group compression, IA adaptation, copy/localization adjustment, interaction-state coverage, design-system pass, or selected-screen regeneration.",
+                "Detect conflicts between the tune request, locked elements, product boundaries, non-goals, privacy controls, safety boundaries, and accepted prototype evidence before prompt writing.",
+                "Write MUST_INHERIT, MUST_ADD, MUST_REMOVE, and FLEXIBLE_CHANGE rules globally and per target screen.",
+                "Build a screen_delta_matrix that binds each target screen to source screen ids, preserve/add/remove/transform/flexible rules, and acceptance criteria.",
+                "Bind every refined prompt to target screen id, source screen id(s), generation scope, target form factor, negative constraints, and acceptance criteria.",
+                "Write generation_order and acceptance_checklist so downstream generation agents can run the refined prompt pack without guessing.",
                 "Record decision audit outcome as revise, continue, pivot, stop, or needs_more_evidence.",
             ],
             after: [
-                "Write updated prototype evidence and review artifacts.",
+                "Write REFINED_PROTO_PROMPT_PACK.yaml, REFINED_PROTO_PROMPT_PACK.md, REVIEW_PLAN.md, and compact EVIDENCE.yaml.",
                 "Write or update the internal decision audit record.",
+                "Refresh prototype SUMMARY.yaml and CURRENT_STATE.yaml after the revision outcome is known.",
                 "Show the user only the tuning result, unresolved question if any, and the next user-facing command.",
             ],
         },
         antiPatterns: [
             "Do not ask the user to manually invoke /ow:decision during a tune loop.",
-            "Do not restart full prototype discovery when a focused revision is enough.",
+            "Do not restart full strategic prototype discovery when a focused refinement is enough.",
+            "Do not tune from one representative screen when the input is a screen group unless the user explicitly limits scope.",
+            "Do not silently drop accepted baseline controls, privacy affordances, memory controls, or non-goals.",
+            "Do not generate HTML, CSS, or runnable app work from /ow:tune.",
             "Do not create design, specs, changes, or runtime work from unaccepted tune evidence.",
-            "Do not tune outside the current validation scope unless the user explicitly changes the target or validation.",
+            "Do not ignore the current validation scope when one is explicitly present and accepted for the prototype.",
         ],
         internalSections: [
             {
                 tag: "target_resolution",
                 items: [
-                    "/ow:tune resolves to the current prototype by default.",
+                    "/ow:tune resolves to the latest approved prototype prompt pack, refined prompt pack, or accepted baseline screen group by default.",
                     "/ow:tune:proto is an explicit alias for tuning the current prototype.",
-                    "/ow:tune:<target> reserves routing for explicit future artifact targets; in M17, implement prototype target behavior and record unsupported targets as unresolved.",
+                    "If no baseline prototype exists, return to /ow:proto instead of inventing refinement context.",
+                ],
+            },
+            {
+                tag: "multi_round_baseline_inheritance",
+                items: [
+                    "Before baseline audit, write baseline_resolution with latest_approved_baseline_group_id, latest_approved_baseline_ref, baseline_lineage, resolution_rule, and stale_source_guard.",
+                    "Use the latest approved baseline group by default; only use an older source screen group when the user explicitly names it.",
+                    "Write carry_forward with locked_screens, locked_elements, preserved_improvements, explicit_unlocks, and cumulative_drift_guard.",
+                    "Carry forward accepted improvements and locked elements across tune rounds unless the tune request explicitly unlocks or removes them.",
+                    "Never silently regenerate from stale source screens when a newer accepted tune pass exists.",
+                ],
+            },
+            {
+                tag: "input_normalization",
+                items: [
+                    "Normalize baseline_source_type, baseline_refs, tune_request, target_form_factor, regeneration_scope, target_screen_count, locked_screens, locked_elements, output_language, target_tool, and constraints before audit.",
+                    "Assign source ids such as SRC_M01, SRC_W01, or SRC_01 when baseline screens lack stable ids.",
+                    "Assign target ids such as MOB_S01, WEB_S01, TAB_S01, or VAR_A_S01 before writing screen-bound prompts.",
+                ],
+            },
+            {
+                tag: "baseline_screen_audit",
+                items: [
+                    "For each source screen, record screen id, screen name, journey stage, user goal, system state, components, copy tone, represented feature, AI/system behavior, trust controls, visual cues, must-preserve elements, platform artifacts to transform or remove, and assumptions.",
+                    "Treat the screen group as one product system, not unrelated images.",
+                    "State excluded source screens explicitly when the user limits scope.",
+                ],
+            },
+            {
+                tag: "product_system_extraction",
+                items: [
+                    "Extract product thesis, target user, core behavior, primary product loop, brand promise, interaction model, information architecture, design language, component vocabulary, copywriting style, feature system, trust and boundary system, and anti-goals.",
+                    "Separate stable_constants from adaptable_variables before interpreting the tune request.",
+                    "Preserve product meaning and interaction logic over raw geometry when adapting form factors.",
+                ],
+            },
+            {
+                tag: "tune_request_interpretation",
+                items: [
+                    "Classify the request into visual refinement, brand alignment, screen-specific correction, feature addition, feature removal, form-factor transformation, screen group expansion, screen group compression, layout or IA adaptation, copy/localization adjustment, interaction-state coverage, design-system pass, or selected-screen regeneration.",
+                    "Apply conflict priority in this order: latest explicit user request, hard constraints, product vision/non-goals/privacy/safety/trust boundaries, locked screens or elements, baseline product system, original upstream prompt, then agent aesthetic judgment.",
+                    "If a request conflicts with a product boundary or locked constraint, record the conflict and generate the closest valid alternative instead of drifting silently.",
+                ],
+            },
+            {
+                tag: "inheritance_delta_rules",
+                items: [
+                    "Build MUST_INHERIT, MUST_ADD, MUST_REMOVE, and FLEXIBLE_CHANGE buckets before writing prompts.",
+                    "Requested removals must appear in global negative constraints, per-screen negative constraints, and acceptance checks.",
+                    "Flexible changes must remain inside the product thesis, brand promise, non-goals, and screen purpose.",
+                    "Build screen_delta_matrix rows with target screen id, source screen ids, preserve, add, remove, transform, flexible, and acceptance criteria.",
                 ],
             },
             {
-                tag: "proto_orchestration",
+                tag: "screen_manifest",
                 items: [
-                    "When no current prototype exists but a current validation target exists, use /ow:proto behavior to create the first prototype evidence before tuning.",
-                    "When a current prototype exists, revise it in place unless the user explicitly requests a new prototype branch.",
-                    "Keep the revision scoped to the user's feedback and the active validation question.",
+                    "Preserve target screen ids across rounds unless screens are deleted, split, merged, or explicitly renamed.",
+                    "Every screen prompt must include prompt_id, target_screen_id, screen_name, source_screen_ids, target form factor, generation scope, dependencies, prompt, negative prompt, and acceptance criteria.",
+                    "Do not output anonymous prompts that downstream generation cannot map back to screens.",
                 ],
             },
             {
-                tag: "revision_protocol",
+                tag: "refined_prompt_pack_output",
                 items: [
-                    "Treat user feedback as the tune brief; ask one clarifying question only when the requested revision is ambiguous or unsafe.",
-                    "Update the smallest artifact set needed: prototype evidence, note, review surface, and evidence files.",
-                    "Preserve visual concept policy, evidence refs, verification, and self-critique integrity from the prototype evidence contract.",
+                    "Output Baseline System Summary, Tuning Intent Summary, Inheritance and Delta Rules, Screen Mapping / Prompt Manifest, Global Design System Prompt, Screen-Specific Refined Prompts, Negative Prompt / Exclusion Rules, Generation Order, and Acceptance Checklist.",
+                    "Every screen-specific refined prompt must be standalone and bound to a target screen id plus source screen ids.",
+                    "Output prompt text only unless a later internal generation stage is explicitly selected; /ow:tune itself does not generate images.",
                 ],
             },
             {
@@ -421,7 +1440,13 @@ function decisionProtocol() {
         auditCheckpoints: {
             before: ["Confirm prototype evidence exists.", "Load only prototype evidence, user feedback summary, and decision index context."],
             during: ["Record audit outcome as continue, revise, pivot, stop, or needs_more_evidence.", "Keep only decision-rich evidence."],
-            after: ["Write the decision record.", "Authorize /ow:design only when outcome is continue.", "Return control to the user-facing proto or tune command."],
+            after: [
+                "Write the decision record.",
+                "Update CURRENT_STATE.yaml last_decision and next_command.",
+                "Set prototype status to accepted only when outcome is continue, revise_requested when outcome is revise, or superseded when pivoted.",
+                "Authorize /ow:design only when outcome is continue.",
+                "Return control to the user-facing proto or tune command.",
+            ],
         },
         antiPatterns: [
             "Do not infer acceptance without user review or explicit evidence.",
@@ -485,6 +1510,7 @@ function designProtocol() {
             ],
             after: [
                 "Write PRODUCT_DESIGN.yaml only after enough design meaning is stable.",
+                "Refresh design SUMMARY.yaml when summary_policy is configured and update CURRENT_STATE.yaml with current_design and spec readiness.",
                 "Hand off to /ow:spec only when spec readiness is true and blockers are explicit.",
             ],
         },
@@ -536,4 +1562,257 @@ function designProtocol() {
         handoffCommands: ["/ow:spec", "/ow:tune", "/ow:validation"],
     };
 }
+function specProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "accepted-design-to-production-spec",
+        requiredContext: [
+            ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
+            ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
+            ".openworkflow/design/DESIGN_INDEX.yaml",
+        ],
+        optionalContext: [
+            ".openworkflow/design/**/PRODUCT_DESIGN.yaml",
+            ".openworkflow/design/**/TECH_SPEC.yaml",
+            ".openworkflow/design/**/FRONTEND_SPEC.yaml",
+            ".openworkflow/design/**/BACKEND_SPEC.yaml",
+            ".openworkflow/design/**/API_CONTRACT.yaml",
+            ".openworkflow/design/**/DB_SCHEMA_MODEL.yaml",
+            ".openworkflow/specs/SPEC_INDEX.yaml",
+            "AGENT.md",
+            "package.json",
+        ],
+        forbiddenContext: [".openworkflow/runtime/**", ".openworkflow/changes/**"],
+        allowedOutputs: [
+            ".openworkflow/specs/SPEC_INDEX.yaml",
+            ".openworkflow/specs/<id>/SPEC.yaml",
+            ".openworkflow/specs/<id>/NOTE.md",
+        ],
+        forbiddenOutputs: [".openworkflow/changes/**", ".openworkflow/runtime/**"],
+        auditCheckpoints: {
+            before: [
+                "Confirm a product design exists and spec_readiness.ready is true.",
+                "Load only the current design and any conditional design packets named by that design.",
+                "Lazy-create the specs index and spec artifact only when /ow:spec is invoked.",
+            ],
+            during: [
+                "Translate accepted product behavior into one implementable production slice.",
+                "Preserve traceability to design evidence, accepted scope, and known blockers.",
+                "Separate user-facing requirements, technical constraints, interfaces, acceptance criteria, and verification plan.",
+            ],
+            after: [
+                "Write the spec artifact and update SPEC_INDEX.yaml.",
+                "Refresh spec SUMMARY.yaml when summary_policy is configured and update CURRENT_STATE.yaml with current_spec and change readiness.",
+                "Hand off to /ow:change only when the spec has bounded scope, acceptance, and verification.",
+                "Confirm no change or runtime artifacts were created.",
+            ],
+        },
+        antiPatterns: [
+            "Do not create specs from unaccepted or unready product design.",
+            "Do not turn a broad product design into a multi-feature implementation plan.",
+            "Do not create change or runtime artifacts during spec work.",
+            "Do not precreate spec artifacts during init or sync; create them only on /ow:spec.",
+        ],
+        internalSections: [
+            {
+                tag: "lazy_create",
+                items: [
+                    "OpenWorkflow init is minimal: it creates only workflow, audit, and config files.",
+                    "If .openworkflow/specs/ or SPEC_INDEX.yaml is absent, create it during /ow:spec from ARTIFACT_CONTRACTS.yaml.",
+                    "Do not create unrelated stage directories, templates, changes, or runtime files while authoring the spec.",
+                ],
+            },
+            {
+                tag: "spec_quality_bar",
+                items: [
+                    "A production spec must be enough for an implementation agent to work without rereading the full discovery history.",
+                    "Name the exact scope, affected user behavior, interface contracts, non-goals, acceptance checks, verification commands, and risks.",
+                    "Keep rationale compact and reference source artifacts by path instead of copying long evidence.",
+                ],
+            },
+            {
+                tag: "readiness_gate",
+                items: [
+                    "Do not hand off to /ow:change until implementation scope, acceptance, and test plan are explicit.",
+                    "If design packets are missing or contradictory, ask one focused question or hand back to /ow:design.",
+                ],
+            },
+        ],
+        handoffCommands: ["/ow:change", "/ow:design"],
+    };
+}
+function changeProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "production-change-planning",
+        requiredContext: [
+            ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
+            ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
+            ".openworkflow/specs/SPEC_INDEX.yaml",
+        ],
+        optionalContext: [
+            ".openworkflow/specs/**/SPEC.yaml",
+            ".openworkflow/specs/**/NOTE.md",
+            ".openworkflow/changes/CHANGE_INDEX.yaml",
+            "AGENT.md",
+            "package.json",
+        ],
+        forbiddenContext: [".openworkflow/runtime/**"],
+        allowedOutputs: [
+            ".openworkflow/changes/CHANGE_INDEX.yaml",
+            ".openworkflow/changes/<id>/CHANGE.yaml",
+            ".openworkflow/changes/<id>/NOTE.md",
+            ".openworkflow/changes/<id>/WORK_ITEMS.yaml",
+        ],
+        forbiddenOutputs: [".openworkflow/runtime/**"],
+        auditCheckpoints: {
+            before: [
+                "Confirm a focused production spec exists.",
+                "Inspect the repository just enough to identify affected paths, integration points, and verification commands.",
+                "Apply coder preflight before finalizing selected implementation boundaries: identify source truth, owned paths, forbidden paths, generated surfaces, and the honest validation ladder.",
+                "Lazy-create the changes index, change artifact, and work items only when /ow:change is invoked.",
+            ],
+            during: [
+                "Convert the spec into one bounded implementation change with non-goals and rollback notes.",
+                "Split work into ordered items with owned paths, dependencies, acceptance, verification, and coder evidence expectations when source edits are likely.",
+                "Record unresolved implementation risks instead of expanding scope.",
+            ],
+            after: [
+                "Write CHANGE.yaml, WORK_ITEMS.yaml, and CHANGE_INDEX.yaml.",
+                "Refresh change SUMMARY.yaml when summary_policy is configured and update CURRENT_STATE.yaml with current_change and runtime readiness.",
+                "Hand off to /ow:team only when work items are implementable, verification is explicit, and coder preflight expectations are visible to the execution agent.",
+                "Confirm no runtime artifacts were created.",
+            ],
+        },
+        antiPatterns: [
+            "Do not implement product code during /ow:change.",
+            "Do not create runtime or agent team files before the change plan is accepted.",
+            "Do not plan work that is not traceable to the current spec.",
+            "Do not precreate change artifacts during init or sync; create them only on /ow:change.",
+        ],
+        internalSections: [
+            {
+                tag: "lazy_create",
+                items: [
+                    "OpenWorkflow init is minimal and does not create .openworkflow/changes/.",
+                    "If CHANGE_INDEX.yaml is absent, create it together with the first change artifact during /ow:change.",
+                    "Create WORK_ITEMS.yaml only for the active change, not as a global task backlog.",
+                ],
+            },
+            {
+                tag: "planning_quality_bar",
+                items: [
+                    "A change plan must let an implementation agent start with bounded files, ordered tasks, acceptance checks, and rollback awareness.",
+                    "Prefer small coherent work items with explicit owned_paths and verification over broad task buckets.",
+                    "Keep the user-facing summary short and keep detailed implementation intelligence in the artifacts.",
+                ],
+            },
+            {
+                tag: "coder_preflight",
+                items: [
+                    "Before work items are ready for implementation, apply coder preflight from skills/coder/SKILL.md.",
+                    "Record source truth, derived generated surfaces, owned_paths, forbidden paths, and validation ladder expectations in the change artifacts.",
+                    "For behavior, validator, CLI report, generated-surface, summary, queue, or git-evidence changes, require the implementation agent to plan RED/GREEN evidence or explicitly mark RED not applicable.",
+                    "/ow:change remains planning; coder governance constrains the future source edits and does not execute them.",
+                ],
+            },
+            {
+                tag: "readiness_gate",
+                items: [
+                    "Do not hand off to /ow:team until CHANGE.yaml and WORK_ITEMS.yaml agree on scope and verification.",
+                    "If the spec is too broad or thin, ask one focused question or hand back to /ow:spec.",
+                ],
+            },
+        ],
+        handoffCommands: ["/ow:team", "/ow:spec"],
+    };
+}
+function teamProtocol() {
+    return {
+        depth: "deep",
+        interactionMode: "approved-change-team-execution",
+        requiredContext: [
+            ".openworkflow/workflow/WORKFLOW_INDEX.yaml",
+            ".openworkflow/audit/ARTIFACT_CONTRACTS.yaml",
+            ".openworkflow/changes/CHANGE_INDEX.yaml",
+        ],
+        optionalContext: [
+            ".openworkflow/changes/**/CHANGE.yaml",
+            ".openworkflow/changes/**/WORK_ITEMS.yaml",
+            ".openworkflow/runtime/RUNTIME_INDEX.yaml",
+            ".openworkflow/runtime/**/STATE.yaml",
+            "AGENT.md",
+            "package.json",
+        ],
+        forbiddenContext: [],
+        allowedOutputs: [
+            ".openworkflow/runtime/RUNTIME_INDEX.yaml",
+            ".openworkflow/runtime/<id>/STATE.yaml",
+            ".openworkflow/runtime/<id>/NOTE.md",
+            ".openworkflow/runtime/<id>/ISSUES.yaml",
+            ".openworkflow/runtime/<id>/CHECKPOINTS.yaml",
+        ],
+        forbiddenOutputs: [],
+        auditCheckpoints: {
+            before: [
+                "Confirm an approved or active change plan and work items exist.",
+                "Audit git status, relevant source files, and any existing runtime state before execution.",
+                "Recover coder governance expectations from the change plan before source edits: owner map, RED/GREEN target, self-check, validation ladder, and evidence binding.",
+                "Lazy-create runtime state only when /ow:team is invoked for an approved change.",
+            ],
+            during: [
+                "Execute work items in dependency order and keep runtime state current.",
+                "Delegate only when the task can run independently with clear owned paths and acceptance.",
+                "Preserve RED/GREEN evidence, post-write self-check notes, verification results, checkpoints, and residual risks as development proceeds.",
+            ],
+            after: [
+                "Update runtime state, issues, and checkpoints.",
+                "Refresh runtime SUMMARY.yaml when summary_policy is configured and update CURRENT_STATE.yaml with current_run, blockers, and next action.",
+                "Run the verification named by the change plan when practical.",
+                "Report changed artifacts, coder evidence status, verification result, and remaining blockers.",
+            ],
+        },
+        antiPatterns: [
+            "Do not start runtime work without a current change plan and work items.",
+            "Do not create team runtime during init, sync, spec, or change planning.",
+            "Do not leave delegated work without ownership, acceptance, or status.",
+            "Do not hide failed verification; record it in runtime issues or checkpoints.",
+        ],
+        internalSections: [
+            {
+                tag: "lazy_create",
+                items: [
+                    "OpenWorkflow init is minimal and does not create .openworkflow/runtime/.",
+                    "Create RUNTIME_INDEX.yaml and the first runtime state only when /ow:team begins approved execution.",
+                    "If runtime already exists, reconcile it instead of replacing historical state.",
+                ],
+            },
+            {
+                tag: "execution_quality_bar",
+                items: [
+                    "Team runtime must preserve enough state for another agent to continue without reading the full conversation.",
+                    "Track active change, active work item, assigned owner or agent, status, verification, issues, and checkpoints.",
+                    "Keep implementation and QA evidence linked to the change plan.",
+                ],
+            },
+            {
+                tag: "coder_execution_gate",
+                items: [
+                    "For source-edit work, follow skills/coder/SKILL.md before completion: owner/file/dependency map, RED evidence when applicable, GREEN evidence after edits, and post-write self-check.",
+                    "Bind coder validation evidence to runtime checkpoints, issues, or the selected-change LOCAL_COMMIT_EVIDENCE.yaml when implementation files changed.",
+                    "Missing coder evidence is guidance at this stage, not a hard failure; record the reason when RED evidence is not applicable.",
+                    "/ow:team remains execution governance and must not turn /ow:coder into a user-facing command.",
+                ],
+            },
+            {
+                tag: "handoff_gate",
+                items: [
+                    "When work is incomplete, leave the next action and blocker explicit in runtime state.",
+                    "When work is complete, record verification and checkpoint readiness before final handoff.",
+                ],
+            },
+        ],
+        handoffCommands: ["/ow:change"],
+    };
+}
 //# sourceMappingURL=registry.js.map