@mmerterden/multi-agent-pipeline 18.0.0 → 19.1.0

package/pipeline/schemas/prefs.schema.json

@@ -14,8 +14,8 @@
   "properties": {
     "schemaVersion": {
       "type": "string",
-      "enum": ["2.0.0", "2.1.0", "2.2.0", "2.3.0", "2.4.0", "2.5.0", "2.6.0"],
-      "description": "v2.0.0: pre-v3.7. v2.1.0: v3.7+ adds identities[].servicePatMap, platformIdentityRouting, recentGroups, recentBranches, serviceStatus, settings, expanded keychainMapping. v2.2.0: v6.0.0 formalizes v5.7 / v5.8 additions (reportChannels, reportContent with technicalAnalysis, wikiScope, autopilotReportTimeoutSeconds) that had been running as 2.1.0 sub-migrations without a proper version bump. v2.5.0: v14.0.0 adds global.skillConformance (Phase 4 criteria resolution) and declares global.ship.autoFix, which the tail command's spec had referenced as global.finish.autoFix without ever declaring it. v2.6.0: v15.0.0 renames global.ship to global.resumeLocal (/multi-agent:ship -> :resume-local)."
+      "enum": ["2.0.0", "2.1.0", "2.2.0", "2.3.0", "2.4.0", "2.5.0", "2.6.0", "2.7.0"],
+      "description": "v2.0.0: pre-v3.7. v2.1.0: v3.7+ adds identities[].servicePatMap, platformIdentityRouting, recentGroups, recentBranches, serviceStatus, settings, expanded keychainMapping. v2.2.0: v6.0.0 formalizes v5.7 / v5.8 additions (reportChannels, reportContent with technicalAnalysis, wikiScope, autopilotReportTimeoutSeconds) that had been running as 2.1.0 sub-migrations without a proper version bump. v2.5.0: v14.0.0 adds global.skillConformance (Phase 4 criteria resolution) and declares global.ship.autoFix, which the tail command's spec had referenced as global.finish.autoFix without ever declaring it. v2.6.0: v15.0.0 renames global.ship to global.resumeLocal (/multi-agent:ship -> :resume-local). v2.7.0: v19.0.0 drops the analysis Lite mode (global.analysisPhase.mode keeps a single value 'full') and adds global.modelRouting, which ships disabled."
     },
     "global": {
       "type": "object",
@@ -369,7 +369,7 @@
         },
         "multiRepoIntegrationHosts": {
           "type": "array",
-          "description": "v5.6.0+. Learn-once registry of host projects that build together with a multi-repo combo (codegen producer → consumer + host integration project). Pipeline checks this registry in Phase 6 before commit; on match, auto-runs the host build. On miss for a ≥2-repo task, prompts the user once and persists the answer. See refs/multi-repo-integration-build.md for the full contract.",
+          "description": "v5.6.0+. Learn-once registry of host projects that build together with a multi-repo combo (codegen producer → consumer + host integration project). Pipeline checks this registry in Phase 4 before commit; on match, auto-runs the host build. On miss for a ≥2-repo task, prompts the user once and persists the answer. See refs/multi-repo-integration-build.md for the full contract.",
           "items": {
             "type": "object",
             "additionalProperties": false,
@@ -470,14 +470,14 @@
               "minimum": 1,
               "maximum": 20,
               "default": 5,
-              "description": "Phase 6 push-must-succeed rebase-retry count before prompting user."
+              "description": "Phase 4 push-must-succeed rebase-retry count before prompting user."
             },
             "buildRetryMax": {
               "type": "integer",
               "minimum": 1,
               "maximum": 10,
               "default": 3,
-              "description": "Phase 3 build retry count before prompting user."
+              "description": "Phase 2 build retry count before prompting user."
             },
             "serviceStatusCacheSeconds": {
               "type": "integer",
@@ -506,12 +506,12 @@
             "pushMustSucceed": {
               "type": "boolean",
               "default": true,
-              "description": "Phase 6 policy. When true, commit + push with rebase-retry up to pushRetryMax. When false, single push attempt, pause on fail."
+              "description": "Phase 4 policy. When true, commit + push with rebase-retry up to pushRetryMax. When false, single push attempt, pause on fail."
             },
             "worktreeAutoRemoveOnPr": {
               "type": "boolean",
               "default": true,
-              "description": "v14.1.0+ - Phase 6 policy. When true, a task's worktree is removed once its PR is open: artefacts (agent-state, triage output, .pipeline/, build+test logs) are salvaged into the log dir first, then `git worktree remove` runs. The BRANCH is kept and is NOT checked out, so the user's own HEAD and uncommitted work are untouched. Default true because the complaint it answers is forgetting to clean up, and every destructive path is gated: a worktree with real uncommitted changes, an unpushed HEAD, --local mode, or a cwd inside the worktree all skip with a reason. Set false to keep worktrees until /multi-agent:kill or :garbage-collect."
+              "description": "v14.1.0+ - Phase 4 policy. When true, a task's worktree is removed once its PR is open: artefacts (agent-state, triage output, .pipeline/, build+test logs) are salvaged into the log dir first, then `git worktree remove` runs. The BRANCH is kept and is NOT checked out, so the user's own HEAD and uncommitted work are untouched. Default true because the complaint it answers is forgetting to clean up, and every destructive path is gated: a worktree with real uncommitted changes, an unpushed HEAD, --local mode, or a cwd inside the worktree all skip with a reason. Set false to keep worktrees until /multi-agent:kill or :garbage-collect."
             }
           }
         },
@@ -560,7 +560,7 @@
             "fableEnabled": {
               "type": "boolean",
               "default": false,
-              "description": "Whether the fable rung is available at all on Claude Code. Ships OFF: the top rung is the most expensive thing a run can reach, and a cost control that is on by default is not a control. false makes every persona that declares preferredModel: fable dispatch on opus from the first call, with no dispatch attempt on fable and no error to recover from  -  a cost control, not a fallback. Claude Code only: Copilot CLI does not offer Fable 5, and on Codex CLI the fable rung means gpt-5.6 @ xhigh, which this switch deliberately does not touch. Turning it off also collapses the Phase 4 Claude Code reviewer panel from three to two (Reviewer 1 lands on opus, which Reviewer 2 already holds), and consensus.reviewerCount records 2."
+              "description": "Whether the fable rung is available at all on Claude Code. Ships OFF: the top rung is the most expensive thing a run can reach, and a cost control that is on by default is not a control. false makes every persona that declares preferredModel: fable dispatch on opus from the first call, with no dispatch attempt on fable and no error to recover from  -  a cost control, not a fallback. Claude Code only: Copilot CLI does not offer Fable 5, and on Codex CLI the fable rung means gpt-5.6 @ xhigh, which this switch deliberately does not touch. Turning it off also collapses the Phase 3 Claude Code reviewer panel from three to two (Reviewer 1 lands on opus, which Reviewer 2 already holds), and consensus.reviewerCount records 2."
             },
             "onDispatchError": {
               "type": "boolean",
@@ -569,6 +569,50 @@
             }
           }
         },
+        "modelRouting": {
+          "type": "object",
+          "description": "Policy-driven model selection. Ships DISABLED and changes no dispatch behaviour while enabled is false. Full contract and the reason `scope` has no host-session member: pipeline/schemas/route-config.schema.json. Written by /multi-agent:route-on, cleared of its on-state (not its rules) by /multi-agent:route-off, reported by /multi-agent:route-status.",
+          "additionalProperties": false,
+          "required": ["enabled"],
+          "properties": {
+            "enabled": {
+              "type": "boolean",
+              "default": false,
+              "description": "Master switch. Ships false."
+            },
+            "strategy": {
+              "type": "string",
+              "enum": ["manual", "task-fit", "cost-ceiling"],
+              "default": "manual"
+            },
+            "scope": {
+              "type": "array",
+              "default": ["subagent"],
+              "uniqueItems": true,
+              "items": {
+                "type": "string",
+                "enum": ["subagent", "bulk-read", "research"]
+              },
+              "description": "No host-session member, deliberately: rewriting the host's base URL would route the user's whole session, not just this pipeline's calls."
+            },
+            "rules": {
+              "type": "array",
+              "default": [],
+              "items": {
+                "type": "object"
+              }
+            },
+            "budgetCeilingUsd": {
+              "type": ["number", "null"],
+              "default": null,
+              "minimum": 0
+            },
+            "recordDecisions": {
+              "type": "boolean",
+              "default": true
+            }
+          }
+        },
         "hosts": {
           "type": "object",
           "additionalProperties": false,
@@ -665,12 +709,12 @@
         "fortify": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v15.14+ - Fortify SSC behaviour. The host and the API token live in global.hosts.fortify and global.keychainMapping.fortify; this object holds the two things that are neither. Phase 4 Gate 5 read `fortify.alwaysCheck` from the moment it shipped, but global was closed to additional properties and this object did not exist, so setting it failed validation and the gate could only ever run off a referenced URL.",
+          "description": "v15.14+ - Fortify SSC behaviour. The host and the API token live in global.hosts.fortify and global.keychainMapping.fortify; this object holds the two things that are neither. Phase 3 Gate 5 read `fortify.alwaysCheck` from the moment it shipped, but global was closed to additional properties and this object did not exist, so setting it failed validation and the gate could only ever run off a referenced URL.",
           "properties": {
             "alwaysCheck": {
               "type": "boolean",
               "default": false,
-              "description": "Run the Phase 4 Fortify gate on every run, not only when the task referenced a finding. Needs versionIds to know what to scan."
+              "description": "Run the Phase 3 Fortify gate on every run, not only when the task referenced a finding. Needs versionIds to know what to scan."
             },
             "versionIds": {
               "type": "array",
@@ -685,7 +729,7 @@
         "reportChannels": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v5.7+ - Phase 7 / /multi-agent:channels kanal seçimi default'ları. Multi-select menüde tick'li gelecek kanallar. Her kanal bağımsız boolean. Autopilot Phase 7'de ALWAYS pauses (30-min timeout) - bu değerler sadece menünün önceden seçili halini belirler.",
+          "description": "v5.7+ - Phase 5 / /multi-agent:channels kanal seçimi default'ları. Multi-select menüde tick'li gelecek kanallar. Her kanal bağımsız boolean. Autopilot Phase 5'de ALWAYS pauses (30-min timeout) - bu değerler sadece menünün önceden seçili halini belirler.",
           "properties": {
             "pr": {
               "type": "boolean",
@@ -718,7 +762,7 @@
         "reportContent": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v5.7+ - Phase 7 / /multi-agent:channels içerik seçimi default'ları. Multi-select menüde tick'li gelecek content source'ları.",
+          "description": "v5.7+ - Phase 5 / /multi-agent:channels içerik seçimi default'ları. Multi-select menüde tick'li gelecek content source'ları.",
           "properties": {
             "normalAnalysis": {
               "type": "boolean",
@@ -728,7 +772,7 @@
             "technicalAnalysis": {
               "type": "boolean",
               "default": false,
-              "description": "Changes (değişen dosyalar gruplanıp ne/neden), Architecture (structural decisions), Dependencies (yeni import/framework/paket). PR body'deki 'Technical Details' bölümünün özeti; user'ın kanal seçimi PR içermediği durumlarda (ör. sadece Jira/Confluence) teknik içerik aktarmak istiyorsa devreye girer. Source: Phase 2 planning + Phase 3 dev log + PR diff stat."
+              "description": "Changes (değişen dosyalar gruplanıp ne/neden), Architecture (structural decisions), Dependencies (yeni import/framework/paket). PR body'deki 'Technical Details' bölümünün özeti; user'ın kanal seçimi PR içermediği durumlarda (ör. sadece Jira/Confluence) teknik içerik aktarmak istiyorsa devreye girer. Source: Phase 1 planning + Phase 2 dev log + PR diff stat."
             },
             "testScenarios": {
               "type": "boolean",
@@ -753,7 +797,7 @@
             "workSummary": {
               "type": "boolean",
               "default": false,
-              "description": "v7.1.0+ - Executive 'Work Done' summary block. Distills the whole pipeline run into a single-screen section: task + branch + base + PR number, scope delivered (✅/⏳ per Phase 2 task), changed files with +/- counts (capped at 20 rows), review outcome (accepted/deferred/rejected counts + approved flag), and a one-line phase tick strip (0 Init ✅ · 1 Analysis ✅ · ...). Source: `agent-state.json` + `phase-tracker.json` + `git diff --numstat` between `baseBranch`...HEAD. Consumed by `render-work-summary.sh`. Greyed out if no state file exists for the task. Opt-in - off by default so baseline PR body stays unchanged."
+              "description": "v7.1.0+ - Executive 'Work Done' summary block. Distills the whole pipeline run into a single-screen section: task + branch + base + PR number, scope delivered (✅/⏳ per Phase 1 task), changed files with +/- counts (capped at 20 rows), review outcome (accepted/deferred/rejected counts + approved flag), and a one-line phase tick strip (0 Init ✅ · 1 Analysis ✅ · ...). Source: `agent-state.json` + `phase-tracker.json` + `git diff --numstat` between `baseBranch`...HEAD. Consumed by `render-work-summary.sh`. Greyed out if no state file exists for the task. Opt-in - off by default so baseline PR body stays unchanged."
             }
           },
           "default": {
@@ -771,7 +815,7 @@
           "default": 1800,
           "minimum": 60,
           "maximum": 7200,
-          "description": "v5.7+ - Phase 7'de autopilot always-pause menüsünde kullanıcı cevap vermezse session'ı sonlandırma süresi (saniye). Default 1800 (30 dk). Timeout'ta external delivery aborted, internal capture (agent-log, telemetry, knowledge) yine çalışır, session /multi-agent:resume ile devam ettirilebilir."
+          "description": "v5.7+ - Phase 5'de autopilot always-pause menüsünde kullanıcı cevap vermezse session'ı sonlandırma süresi (saniye). Default 1800 (30 dk). Timeout'ta external delivery aborted, internal capture (agent-log, telemetry, knowledge) yine çalışır, session /multi-agent:resume ile devam ettirilebilir."
         },
         "wikiScope": {
           "type": "array",
@@ -814,7 +858,7 @@
         "triageCrossCheck": {
           "type": "object",
           "additionalProperties": false,
-          "description": "Optional second-opinion check on Phase 4 triage decisions. When enabled, a configurable percentage of triage outputs are re-classified by Sonnet and the diff vs. Opus's verdict is logged for audit. Mitigates the single-point-of-failure risk that Opus triage hallucinates or drifts. Adds latency + cost; default off.",
+          "description": "Optional second-opinion check on Phase 3 triage decisions. When enabled, a configurable percentage of triage outputs are re-classified by Sonnet and the diff vs. Opus's verdict is logged for audit. Mitigates the single-point-of-failure risk that Opus triage hallucinates or drifts. Adds latency + cost; default off.",
           "properties": {
             "enabled": {
               "type": "boolean",
@@ -837,19 +881,19 @@
             "blockOnDisagreement": {
               "type": "boolean",
               "default": false,
-              "description": "If true, a cross-check disagreement pauses Phase 5/6 for human review. If false (default), the disagreement is logged to metrics.jsonl as `triage.cross_check_diff` and the pipeline proceeds with Opus's original verdict."
+              "description": "If true, a cross-check disagreement pauses Phase 3/4 for human review. If false (default), the disagreement is logged to metrics.jsonl as `triage.cross_check_diff` and the pipeline proceeds with Opus's original verdict."
             }
           }
         },
         "testBaseline": {
           "type": "object",
           "additionalProperties": false,
-          "description": "Phase 0 test baseline. When enabled, Phase 0 runs the SAME test command Phase 4 Gate 3 uses and records which tests were already failing before this run touched anything, so Phase 4 stops attributing an inherited red suite to the current work. Three outcomes are stored, never two: green, red (with the failing set when it can be parsed, otherwise the log path alone), or unknown when the command is absent or the time cap was hit. Off by default because on iOS the run costs a full xcodebuild test before any work starts. Pattern source: obra/superpowers using-git-worktrees Step 3 'Verify Clean Baseline', extended from ask-the-user to a stored set Phase 4 can subtract.",
+          "description": "Phase 0 test baseline. When enabled, Phase 0 runs the SAME test command Phase 3 Gate 3 uses and records which tests were already failing before this run touched anything, so Phase 3 stops attributing an inherited red suite to the current work. Three outcomes are stored, never two: green, red (with the failing set when it can be parsed, otherwise the log path alone), or unknown when the command is absent or the time cap was hit. Off by default because on iOS the run costs a full xcodebuild test before any work starts. Pattern source: obra/superpowers using-git-worktrees Step 3 'Verify Clean Baseline', extended from ask-the-user to a stored set Phase 3 can subtract.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": false,
-              "description": "Master switch. Off by default - a baseline run costs one full test suite before Phase 3 starts."
+              "description": "Master switch. Off by default - a baseline run costs one full test suite before Phase 2 starts."
             },
             "timeoutSeconds": {
               "type": "integer",
@@ -863,11 +907,11 @@
         "reviewDisagreementRound": {
           "type": "boolean",
           "default": false,
-          "description": "v6.1.0+ - Phase 4 Step 2.5 rebuttal round. When reviewers disagree (mixed blocker/approved verdict), each reviewer is re-prompted with the others' opposing arguments for one additional round before triage. Lifts signal quality on ambiguous findings at ~1× Step 2 token cost. Off by default - flip for security-critical or release-branch reviews."
+          "description": "v6.1.0+ - Phase 3 Step 2.5 rebuttal round. When reviewers disagree (mixed blocker/approved verdict), each reviewer is re-prompted with the others' opposing arguments for one additional round before triage. Lifts signal quality on ambiguous findings at ~1× Step 2 token cost. Off by default - flip for security-critical or release-branch reviews."
         },
         "analysisProfiles": {
           "type": "array",
-          "description": "v16.6+ - which analysis standards the /multi-agent:analysis Step 1b picker offers (Locked 32). Listing one value auto-resolves the step instead of asking a question whose answer is already settled. Omitted means both are offered.",
+          "description": "v16.6+ - which analysis standards the /multi-agent:analysis Step 1b picker offers (Locked 31). Listing one value auto-resolves the step instead of asking a question whose answer is already settled. Omitted means both are offered.",
           "items": {
             "type": "string",
             "enum": ["global", "corporate"]
@@ -933,7 +977,7 @@
         "verifyByTest": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v10.8+ - Phase 4 Step 3.7 verify-by-test. When enabled, accepted BLOCKING findings are empirically validated before the Phase 3 rework loop: one verifier agent writes a minimal repro test per finding and runs only that test. Confirmed findings hand their failing test to Phase 3 as the RED step; non-reproducible findings are downgraded to deferred under evidence-gate. Only blocking findings are ever verified (fixed behavior, not a knob). Adds one model call plus up to maxFindings single-test runs per iteration with accepted blockers; default off. Flip on for security-critical work, release branches, or repos with noisy reviewers. Full spec: refs/features/verify-by-test.md.",
+          "description": "v10.8+ - Phase 3 Step 3.7 verify-by-test. When enabled, accepted BLOCKING findings are empirically validated before the Phase 3 rework loop: one verifier agent writes a minimal repro test per finding and runs only that test. Confirmed findings hand their failing test to Phase 3 as the RED step; non-reproducible findings are downgraded to deferred under evidence-gate. Only blocking findings are ever verified (fixed behavior, not a knob). Adds one model call plus up to maxFindings single-test runs per iteration with accepted blockers; default off. Flip on for security-critical work, release branches, or repos with noisy reviewers. Full spec: refs/features/verify-by-test.md.",
           "properties": {
             "enabled": {
               "type": "boolean",
@@ -1035,14 +1079,14 @@
               "minimum": 1,
               "maximum": 3,
               "default": 3,
-              "description": "Phase 4 -> Phase 3 rework cycles before trigger 3 trips. Bounded above by the Phase 3 retryCount hard-kill."
+              "description": "Phase 3 -> Phase 2 rework cycles before trigger 3 trips. Bounded above by the Phase 2 retryCount hard-kill."
             }
           }
         },
         "testStability": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v16.20+ - a test that passes only on retry is a flake signal, not a pass. Phase 3 runs new and changed tests repeatCount times before GREEN; disagreeing outcomes block and are recorded as test.flake_signal.",
+          "description": "v16.20+ - a test that passes only on retry is a flake signal, not a pass. Phase 2 runs new and changed tests repeatCount times before GREEN; disagreeing outcomes block and are recorded as test.flake_signal.",
           "properties": {
             "repeatCount": {
               "type": "integer",
@@ -1056,7 +1100,7 @@
         "review": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.6+  -  Standalone /multi-agent:review knobs. Only affects the post-pr-review.sh path (PR-mode comment posting); does not change Phase 4 reviewer dispatch inside the full pipeline.",
+          "description": "v8.6+  -  Standalone /multi-agent:review knobs. Only affects the post-pr-review.sh path (PR-mode comment posting); does not change Phase 3 reviewer dispatch inside the full pipeline.",
           "properties": {
             "dedupeInlineComments": {
               "type": "boolean",
@@ -1068,7 +1112,7 @@
         "shadowGit": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.6+  -  Cline-style per-tool-call shadow Git checkpoints. A separate git repo under ~/.claude/state/shadow-git/<task-id>/.git/ snapshots the worktree at each Phase 3 step boundary (per Plan Todo) and optionally after each Edit/Write/MultiEdit/Bash mutation. The shadow lives outside the project's real .git so semantic commits stay clean. Pattern source: Cline checkpoints (https://docs.cline.bot/features/checkpoints). Off by default  -  adds a per-step git commit on top of the existing TDD loop; flip on for risky refactors where sub-phase rollback is worth ~50ms per snapshot.",
+          "description": "v8.6+  -  Cline-style per-tool-call shadow Git checkpoints. A separate git repo under ~/.claude/state/shadow-git/<task-id>/.git/ snapshots the worktree at each Phase 2 step boundary (per Plan Todo) and optionally after each Edit/Write/MultiEdit/Bash mutation. The shadow lives outside the project's real .git so semantic commits stay clean. Pattern source: Cline checkpoints (https://docs.cline.bot/features/checkpoints). Off by default  -  adds a per-step git commit on top of the existing TDD loop; flip on for risky refactors where sub-phase rollback is worth ~50ms per snapshot.",
           "properties": {
             "enabled": {
               "type": "boolean",
@@ -1093,12 +1137,12 @@
         "planTodos": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.6+  -  Plan-as-live-Todo-list. Phase 2 emits agent-state.plan.todos[] conforming to pipeline/schemas/plan-todos.schema.json; Phase 3 iterates step-by-step via pipeline/lib/plan-todos.sh next/start/complete; Phase 7 renders the rollup into agent-log.md and the PR body. The plan is broken into a live, structured Todo list. Off by default  -  opt in to add status-transition writes per Phase 3 step in exchange for sub-step visibility + per-step notes.",
+          "description": "v8.6+  -  Plan-as-live-Todo-list. Phase 2 emits agent-state.plan.todos[] conforming to pipeline/schemas/plan-todos.schema.json; Phase 3 iterates step-by-step via pipeline/lib/plan-todos.sh next/start/complete; Phase 5 renders the rollup into agent-log.md and the PR body. The plan is broken into a live, structured Todo list. Off by default  -  opt in to add status-transition writes per Phase 3 step in exchange for sub-step visibility + per-step notes.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": false,
-              "description": "Master switch. When true, Phase 2 Step 4.5 emits plan.todos[] and Phase 3 / 7 read from it."
+              "description": "Master switch. When true, Phase 1 Step 11 emits plan.todos[] and Phase 2 / 5 read from it."
             }
           }
         },
@@ -1221,9 +1265,9 @@
           "properties": {
             "mode": {
               "type": "string",
-              "enum": ["auto", "full", "lite"],
-              "default": "auto",
-              "description": "Section depth. auto picks lite for a bugfix or chore with no Figma reference and full otherwise; full and lite pin it. Layer selection (Analysis / Technical / Development) is a separate axis asked at intake."
+              "enum": ["full"],
+              "default": "full",
+              "description": "Retained with a single value so a preferences file that sets it stays valid. Section depth is no longer a mode: Lite was removed in v19.0.0 because it chose sections from a fixed list while Locked 2 chooses them from evidence, and the two disagreed in both directions. Sections render when they have evidence and are dropped when they do not. The `auto` and `lite` values are rewritten to `full` by the 2.6.0-to-2.7.0 migration."
             },
             "forceFull": {
               "type": "boolean",
@@ -1302,12 +1346,12 @@
         "repoMap": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.6+  -  Aider-style token-budgeted repo map injection. When enabled, the orchestrator runs pipeline/scripts/repo-map.mjs BEFORE Phase 1 (Analysis) and BEFORE Phase 4 (Review) and injects the resulting markdown as ${REPO_MAP} into the reviewer/analyser prompts. Deterministic  -  no embeddings, no network. Pattern source: https://aider.chat/docs/repomap.html. Off by default (introduces a ~150-300ms repo scan + repo-shaped prompt growth); flip on to lower per-call token cost on follow-up exploration phases.",
+          "description": "v8.6+  -  Aider-style token-budgeted repo map injection. When enabled, the orchestrator runs pipeline/scripts/repo-map.mjs BEFORE Phase 1 (Plan) and BEFORE Phase 3 (Review) and injects the resulting markdown as ${REPO_MAP} into the reviewer/analyser prompts. Deterministic  -  no embeddings, no network. Pattern source: https://aider.chat/docs/repomap.html. Off by default (introduces a ~150-300ms repo scan + repo-shaped prompt growth); flip on to lower per-call token cost on follow-up exploration phases.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": false,
-              "description": "Master switch. When true, Phase 1 and Phase 4 receive ${REPO_MAP} as additional context. When false, both phases run with only the diff + their normal prompt shell."
+              "description": "Master switch. When true, Phase 1 and Phase 3 receive ${REPO_MAP} as additional context. When false, both phases run with only the diff + their normal prompt shell."
             },
             "tokenBudget": {
               "type": "integer",
@@ -1336,17 +1380,17 @@
         "codeGraph": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v16.1+  -  deterministic, LLM-free code graph (symbols, imports, references) built by pipeline/scripts/graph-build.mjs into ~/.claude/knowledge/<project>/code-graph.json. Phase 1 Step 2.6 queries it to narrow the Explore starting set; Phase 7 Step 3 refreshes it after the branch changed code. Zero API cost, zero runtime dependencies (ADR-0004), read-only on the repo. Off by default: with it off the pipeline behaves exactly as it did before. Measured on a 4,300-file Swift app at a fixed 30k retrieval budget, it roughly doubled coverage at under half the token cost on domain-word searches, and was marginally worse than plain grep when the task already names an exact symbol.",
+          "description": "v16.1+  -  deterministic, LLM-free code graph (symbols, imports, references) built by pipeline/scripts/graph-build.mjs into ~/.claude/knowledge/<project>/code-graph.json. Phase 1 Step 2.6 queries it to narrow the Explore starting set; Phase 5 Step 3 refreshes it after the branch changed code. Zero API cost, zero runtime dependencies (ADR-0004), read-only on the repo. Off by default: with it off the pipeline behaves exactly as it did before. Measured on a 4,300-file Swift app at a fixed 30k retrieval budget, it roughly doubled coverage at under half the token cost on domain-word searches, and was marginally worse than plain grep when the task already names an exact symbol.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": false,
-              "description": "Master switch. When false, Phase 1 Step 2.6 and the Phase 7 refresh are skipped entirely and no graph file is written."
+              "description": "Master switch. When false, Phase 1 Step 2.6 and the Phase 5 refresh are skipped entirely and no graph file is written."
             },
             "autoRefresh": {
               "type": "boolean",
               "default": true,
-              "description": "Rebuild the graph in Phase 7 after the branch changed code. When false, the graph only refreshes when someone asks for it via /multi-agent:graph."
+              "description": "Rebuild the graph in Phase 5 after the branch changed code. When false, the graph only refreshes when someone asks for it via /multi-agent:graph."
             },
             "maxNodes": {
               "type": "integer",
@@ -1367,12 +1411,12 @@
         "devCritic": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.6+ - Phase 3.5 evaluator-optimizer. After the Dev generator's last edit and BEFORE Phase 4 reviewers, dispatch agents/dev-critic.md (Sonnet by default) to run deterministic gates (build/lint/test/secrets) + the platform checklist (rules/*.md). Max 2 critic iterations, then escalate. Catches gate failures and checklist violations that would otherwise burn 2-3 Phase 4 reviewer calls + Opus triage. Off by default - introduces 1× Sonnet call per Dev iteration; flip on for feature work, security-touching paths, or multi-file refactors. Source: Anthropic 'Building Effective Agents' (Dec 2024) evaluator-optimizer pattern.",
+          "description": "v8.6+ - Phase 2.5 evaluator-optimizer. After the Dev generator's last edit and BEFORE Phase 3 reviewers, dispatch agents/dev-critic.md (Sonnet by default) to run deterministic gates (build/lint/test/secrets) + the platform checklist (rules/*.md). Max 2 critic iterations, then escalate. Catches gate failures and checklist violations that would otherwise burn 2-3 Phase 3 reviewer calls + Opus triage. Off by default - introduces 1× Sonnet call per Dev iteration; flip on for feature work, security-touching paths, or multi-file refactors. Source: Anthropic 'Building Effective Agents' (Dec 2024) evaluator-optimizer pattern.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": false,
-              "description": "Master switch. When true, Phase 3 dispatches dev-critic after the generator's last edit; when false, Phase 3 hands off directly to Phase 4."
+              "description": "Master switch. When true, Phase 2 dispatches dev-critic after the generator's last edit; when false, Phase 2 hands off directly to Phase 3."
             },
             "maxIterations": {
               "type": "integer",
@@ -1398,12 +1442,12 @@
         "diffRiskAdvisory": {
           "type": "boolean",
           "default": true,
-          "description": "v8.3+ - Phase 4 Step 1.75 advisory diff risk scoring. When enabled, `pipeline/scripts/diff-risk-score.mjs` runs before reviewer dispatch and the top-N risk-ranked files are injected into each reviewer's prompt as a priority hint (security paths, public API surfaces, untested source changes, schema migrations). Heuristic, deterministic, no LLM cost. Default ON - the run is sub-second and never gates the pipeline. Flip to false to skip the script and the prompt injection entirely."
+          "description": "v8.3+ - Phase 3 Step 1.75 advisory diff risk scoring. When enabled, `pipeline/scripts/diff-risk-score.mjs` runs before reviewer dispatch and the top-N risk-ranked files are injected into each reviewer's prompt as a priority hint (security paths, public API surfaces, untested source changes, schema migrations). Heuristic, deterministic, no LLM cost. Default ON - the run is sub-second and never gates the pipeline. Flip to false to skip the script and the prompt injection entirely."
         },
         "reviewScopeGate": {
           "type": "boolean",
           "default": true,
-          "description": "v12.8+ - Phase 4 Step 1.77 reviewer-scope gate. Decides the reviewer count from the deterministic diff-risk report via `pipeline/scripts/review-scope.mjs`: a diff under 20 lines of churn with max_score < 3.0 and no security_path / migration / public_api / no_test_change / test_lines_removed signal runs ONE reviewer instead of the full CLI-aware set (2 on Claude Code, 3 on Copilot CLI). No LLM. Fails safe in one direction only - any error, empty report or validator rejection resolves to the full set, because skipping a reviewer trades coverage for cost. Set false to force the full set on every diff.",
+          "description": "v12.8+ - Phase 3 Step 1.77 reviewer-scope gate. Decides the reviewer count from the deterministic diff-risk report via `pipeline/scripts/review-scope.mjs`: a diff under 20 lines of churn with max_score < 3.0 and no security_path / migration / public_api / no_test_change / test_lines_removed signal runs ONE reviewer instead of the full CLI-aware set (2 on Claude Code, 3 on Copilot CLI). No LLM. Fails safe in one direction only - any error, empty report or validator rejection resolves to the full set, because skipping a reviewer trades coverage for cost. Set false to force the full set on every diff.",
           "$comment": "Shipped inert for a release: the script existed, was unit- and smoke-tested, and no phase doc referenced it, so every diff paid for the full reviewer set. Wired in Step 1.77; smoke-gate-wiring.sh keeps it reachable."
         },
         "jiraContext": {
@@ -1435,17 +1479,17 @@
         "priorArtEnrichment": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.3+  -  Per-repo triage memory layer. When enabled, Phase 7 ingests every triage output's accepted/deferred/rejected rows into ~/.claude/memory/multi-agent/<repo-slug>/triage-corpus.jsonl (idempotent), and Phase 4 triage queries the corpus for similar past findings to inject as context. Per-repo isolation  -  never cross-leaks between projects.",
+          "description": "v8.3+  -  Per-repo triage memory layer. When enabled, Phase 5 ingests every triage output's accepted/deferred/rejected rows into ~/.claude/memory/multi-agent/<repo-slug>/triage-corpus.jsonl (idempotent), and Phase 3 triage queries the corpus for similar past findings to inject as context. Per-repo isolation  -  never cross-leaks between projects.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": true,
-              "description": "Master switch for the lookup side. When false, Phase 4 triage runs without prior-art injection. Ingest still runs unless ingestOnComplete is also flipped."
+              "description": "Master switch for the lookup side. When false, Phase 3 triage runs without prior-art injection. Ingest still runs unless ingestOnComplete is also flipped."
             },
             "ingestOnComplete": {
               "type": "boolean",
               "default": true,
-              "description": "When false, Phase 7 skips the corpus append. Use to keep an existing corpus frozen while still benefiting from lookups against it."
+              "description": "When false, Phase 5 skips the corpus append. Use to keep an existing corpus frozen while still benefiting from lookups against it."
             },
             "topN": {
               "type": "integer",
@@ -1464,12 +1508,12 @@
         "learningsLedger": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v9.3+  -  Per-repo persistent LEARNINGS ledger (pipeline/scripts/learnings-ledger.mjs), stored next to triage-corpus at ~/.claude/memory/multi-agent/<repo-slug>/learnings-ledger.jsonl. Holds durable architectural facts, conventions, and explicitly rejected review preferences. A compact brief is injected into Phase 1 analysis and Phase 4 triage so agents stop re-discovering structure and reviewers stop re-flagging rejected feedback (the most-cited cold-boot-amnesia complaint). Phase 7 distills each run's rejected findings into the ledger. Per-repo isolated; never cross-leaks.",
+          "description": "v9.3+  -  Per-repo persistent LEARNINGS ledger (pipeline/scripts/learnings-ledger.mjs), stored next to triage-corpus at ~/.claude/memory/multi-agent/<repo-slug>/learnings-ledger.jsonl. Holds durable architectural facts, conventions, and explicitly rejected review preferences. A compact brief is injected into Phase 1 analysis and Phase 3 triage so agents stop re-discovering structure and reviewers stop re-flagging rejected feedback (the most-cited cold-boot-amnesia complaint). Phase 5 distills each run's rejected findings into the ledger. Per-repo isolated; never cross-leaks.",
           "properties": {
             "enabled": {
               "type": "boolean",
               "default": true,
-              "description": "Master switch. When false, no brief is injected and Phase 7 does not append to the ledger."
+              "description": "Master switch. When false, no brief is injected and Phase 5 does not append to the ledger."
             },
             "injectIntoAnalysis": {
               "type": "boolean",
@@ -1479,7 +1523,7 @@
             "injectIntoTriage": {
               "type": "boolean",
               "default": true,
-              "description": "Inject the rejected-preference brief into Phase 4 triage so known-rejected suggestions are not re-accepted."
+              "description": "Inject the rejected-preference brief into Phase 3 triage so known-rejected suggestions are not re-accepted."
             },
             "maxBriefEntries": {
               "type": "integer",
@@ -1590,7 +1634,7 @@
                 "type": "string"
               },
               "default": ["3"],
-              "description": "Phases the gate never blocks in. Phase 3 is the shipped default and removing it breaks development: Claude Code's Edit requires the same file to have been Read first, so a gate that blocks reads while code is being changed blocks the change. The gate is for phases that read to UNDERSTAND."
+              "description": "Phases the gate never blocks in. Phase 2 is the shipped default and removing it breaks development: Claude Code's Edit requires the same file to have been Read first, so a gate that blocks reads while code is being changed blocks the change. The gate is for phases that read to UNDERSTAND."
             },
             "model": {
               "type": "string",
@@ -1624,7 +1668,7 @@
         "testGap": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v8.3+ - Phase 5 Step 0 advisory test-gap detector. Walks the diff for newly added public symbols and reports those without a paired test (or test method covering them). Heuristic, deterministic, no LLM cost.",
+          "description": "v8.3+ - Phase 3 Step 0 advisory test-gap detector. Walks the diff for newly added public symbols and reports those without a paired test (or test method covering them). Heuristic, deterministic, no LLM cost.",
           "properties": {
             "enabled": {
               "type": "boolean",
@@ -1640,7 +1684,7 @@
               "type": ["integer", "null"],
               "minimum": 1,
               "default": null,
-              "description": "If set, gaps where `important + blocking` count exceeds the threshold are treated as a Phase 4 rework finding (loops back to Phase 3). null = advisory only - Phase 5 prints the report but never gates."
+              "description": "If set, gaps where `important + blocking` count exceeds the threshold are treated as a Phase 3 rework finding (loops back to Phase 2). null = advisory only - Phase 3 prints the report but never gates."
             },
             "promoteSeverity": {
               "type": "boolean",
@@ -1658,12 +1702,12 @@
         "perRepoMemory": {
           "type": "boolean",
           "default": false,
-          "description": "v6.2.0+ - Per-repo file-system memory layer. When enabled, Phase 0 reads `$PROJECT_ROOT/.multi-agent/memory/MEMORY.md` (if present) and injects it into the pipeline context, and Phase 7 dispatches a scoped synthesis subagent that may write new memory files to the same directory. Categories: user / feedback / project / reference (same taxonomy as the conversation-level auto-memory). Memory is local to each repo - never synced to remote, never committed to git (pipeline installer adds `.multi-agent/memory/` to `.gitignore`). Off by default - flip when working on a repo long enough that accumulated project context starts paying for itself."
+          "description": "v6.2.0+ - Per-repo file-system memory layer. When enabled, Phase 0 reads `$PROJECT_ROOT/.multi-agent/memory/MEMORY.md` (if present) and injects it into the pipeline context, and Phase 5 dispatches a scoped synthesis subagent that may write new memory files to the same directory. Categories: user / feedback / project / reference (same taxonomy as the conversation-level auto-memory). Memory is local to each repo - never synced to remote, never committed to git (pipeline installer adds `.multi-agent/memory/` to `.gitignore`). Off by default - flip when working on a repo long enough that accumulated project context starts paying for itself."
         },
         "autopilotSafetyGate": {
           "type": "boolean",
           "default": true,
-          "description": "v7.0.0+ - Phase 2 autopilot safety classifier. Before autopilot mode consumes the user's approval skip, run `classify-plan-safety.mjs` over the approved plan. If the heuristic score ≥ 50 (e.g. >15 files touched, or security-path touch, or delete-without-test, or schema migration) inject a ONE-TIME pause asking for explicit manual approval - even in autopilot. Default ON because the risk of skipping this gate is asymmetric: a pause on a high-blast-radius plan costs seconds; a silent auto-merge of a bad one costs hours of rollback. Flip to `false` only for tightly-scoped autopilot workflows (e.g. batch figma component iteration) where the task class is known-safe."
+          "description": "v7.0.0+ - Phase 1 autopilot safety classifier. Before autopilot mode consumes the user's approval skip, run `classify-plan-safety.mjs` over the approved plan. If the heuristic score ≥ 50 (e.g. >15 files touched, or security-path touch, or delete-without-test, or schema migration) inject a ONE-TIME pause asking for explicit manual approval - even in autopilot. Default ON because the risk of skipping this gate is asymmetric: a pause on a high-blast-radius plan costs seconds; a silent auto-merge of a bad one costs hours of rollback. Flip to `false` only for tightly-scoped autopilot workflows (e.g. batch figma component iteration) where the task class is known-safe."
         },
         "dynamicSkillLoading": {
           "type": "boolean",
@@ -1870,7 +1914,7 @@
         "skillConformance": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v14.0.0+ - Phase 4 Step 1.78 criteria resolution. The stage itself has no on/off key and neither does its exception-expiry check: a run that can switch off its own anti-reward-hacking control cannot be trusted to report a pass (same reasoning as Step 1.76).",
+          "description": "v14.0.0+ - Phase 3 Step 1.78 criteria resolution. The stage itself has no on/off key and neither does its exception-expiry check: a run that can switch off its own anti-reward-hacking control cannot be trusted to report a pass (same reasoning as Step 1.76).",
           "properties": {
             "blockOnCoverageGap": {
               "type": "boolean",
@@ -1899,7 +1943,7 @@
             "enabled": {
               "type": "boolean",
               "default": true,
-              "description": "Off disables capture, upload and both render sections. The Phase 6 blocker goes with it."
+              "description": "Off disables capture, upload and both render sections. The Phase 4 blocker goes with it."
             },
             "maxAttachmentMb": {
               "type": "integer",
@@ -1930,7 +1974,7 @@
         "testDepth": {
           "type": "object",
           "additionalProperties": false,
-          "description": "How far a run tests by default. The question is asked at intake (Phase 0), not in Phase 5, because Phase 5 is not in the phase set for autopilot or either local mode. Autopilot never asks and reads `default` instead; the run records which rule fired.",
+          "description": "How far a run tests by default. The question is asked at intake (Phase 0), not in the user test, because it is not in the step set for autopilot or either local mode. Autopilot never asks and reads `default` instead; the run records which rule fired.",
           "properties": {
             "default": {
               "type": "string",
@@ -2001,7 +2045,7 @@
         "mcpSurface": {
           "type": "object",
           "additionalProperties": false,
-          "description": "v17.6.0+ - how many MCP servers this host has registered. Every registered server's tool list is charged against the context window on EVERY turn, and the user adds them one at a time without ever seeing the running total; our own toolkit contributes 99 tools by itself. `doctor` reports the count, it never disables anything.",
+          "description": "v17.6.0+ - how many MCP servers this host has registered. Every registered server's tool list is charged against the context window on EVERY turn, and the user adds them one at a time without ever seeing the running total; our own toolkit contributes 115 tools by itself. `doctor` reports the count, it never disables anything.",
           "properties": {
             "infoAbove": {
               "type": "integer",
@@ -2024,7 +2068,7 @@
             "testPolicy": {
               "type": "string",
               "enum": ["tdd", "tests-after", "none"],
-              "description": "How Phase 3 authors tests in this project: tdd (default; failing test first), tests-after (implementation first, tests authored at the end), none (no unit/UI test authoring; existing tests are kept and run). Asked once by Phase 0 when absent; autopilot defaults to tdd and notes it."
+              "description": "How Phase 2 authors tests in this project: tdd (default; failing test first), tests-after (implementation first, tests authored at the end), none (no unit/UI test authoring; existing tests are kept and run). Asked once by Phase 0 when absent; autopilot defaults to tdd and notes it."
             },
             "defaultReviewers": {
               "type": "array",
@@ -2098,7 +2142,7 @@
             },
             "componentDevWorkflow": {
               "type": "boolean",
-              "description": "Project follows the component (Configuration / View / Modifiers) development workflow, so Phase 3 dispatches the component skills instead of the standard TDD flow."
+              "description": "Project follows the component (Configuration / View / Modifiers) development workflow, so Phase 2 dispatches the component skills instead of the standard TDD flow."
             },
             "figmaConfigPath": {
               "type": "string",
@@ -2164,7 +2208,7 @@
                 "type": "string"
               },
               "maxItems": 4,
-              "description": "Absolute local paths of this project's counterpart app repo - the same product on the other mobile platform. Read by Phase 4's platform-parity cross-check. Learned, not configured: the first run that resolves a counterpart writes it here and every later run reuses it without asking. An entry whose path is gone, or whose marker files no longer resolve to the mirrored stack, is dropped and re-detected rather than trusted. Same shape as webRoots."
+              "description": "Absolute local paths of this project's counterpart app repo - the same product on the other mobile platform. Read by Phase 3's platform-parity cross-check. Learned, not configured: the first run that resolves a counterpart writes it here and every later run reuses it without asking. An entry whose path is gone, or whose marker files no longer resolve to the mirrored stack, is dropped and re-detected rather than trusted. Same shape as webRoots."
             }
           }
         }

package/pipeline/schemas/reviewer-output.schema.json

@@ -2,8 +2,8 @@
   "$schema": "https://json-schema.org/draft/2020-12/schema",
   "$id": "https://github.com/mmerterden/multi-agent-pipeline/pipeline/schemas/reviewer-output.schema.json",
   "version": "1.3.0",
-  "title": "Multi-Agent Pipeline  -  Phase 4 reviewer output",
-  "description": "Contract for a single code-reviewer subagent's JSON output in Phase 4 Step 2. Every host dispatches 3 parallel reviewers; the middle slot is CLI-aware: Claude Code (Fable, Opus, Sonnet); Copilot CLI (Opus, GPT-5.4, Sonnet); Codex CLI dispatches 3 (gpt-5.6 at xhigh, gpt-5.4, gpt-5.6 at medium). Every reviewer must return an object matching this shape before Opus triage merges them. v1.1.0 adds the rule-ID conformance checklist: when the orchestrator supplies a ${CRITERIA} block (Phase 4 Step 1.78), the reviewer must return one conformance row per selected rule ID. Findings alone cannot answer 'was this applied completely' - a reviewer that opened nothing returns the same empty findings array as one that checked everything. v1.2.0 adds the optional per-finding fingerprint (Phase 4 Step 2.1): the stable id a finding keeps across review rounds.",
+  "title": "Multi-Agent Pipeline  -  Phase 3 reviewer output",
+  "description": "Contract for a single code-reviewer subagent's JSON output in Phase 4 Step 2. Every host dispatches 3 parallel reviewers; the middle slot is CLI-aware: Claude Code (Fable, Opus, Sonnet); Copilot CLI (Opus, GPT-5.4, Sonnet); Codex CLI dispatches 3 (gpt-5.6 at xhigh, gpt-5.4, gpt-5.6 at medium). Every reviewer must return an object matching this shape before Opus triage merges them. v1.1.0 adds the rule-ID conformance checklist: when the orchestrator supplies a ${CRITERIA} block (Phase 3 Step 1.78), the reviewer must return one conformance row per selected rule ID. Findings alone cannot answer 'was this applied completely' - a reviewer that opened nothing returns the same empty findings array as one that checked everything. v1.2.0 adds the optional per-finding fingerprint (Phase 3 Step 2.1): the stable id a finding keeps across review rounds.",
   "type": "object",
   "additionalProperties": false,
   "required": ["findings", "approved"],
@@ -122,7 +122,7 @@
         "criteriaSource": {
           "type": "string",
           "minLength": 1,
-          "description": "Which criteria source the rule came from: a registry name, a module-guide path, or 'exception-marker-audit'. Lets Phase 7 attribute findings to the standard that produced them."
+          "description": "Which criteria source the rule came from: a registry name, a module-guide path, or 'exception-marker-audit'. Lets Phase 5 attribute findings to the standard that produced them."
         },
         "fingerprint": {
           "type": "string",