@glrs-dev/cli 1.0.1 → 1.2.0

package/dist/vendor/harness-opencode/dist/agents/prompts/qa-reviewer.open.md

@@ -0,0 +1,58 @@
+---
+name: qa-reviewer
+description: Fast adversarial reviewer. Always re-runs verifiers. Returns [PASS] or [FAIL]. Default for typical diffs.
+mode: subagent
+model: anthropic/claude-sonnet-4-6
+temperature: 0.1
+---
+
+<!-- STRICT_EXECUTOR_VARIANT -->
+
+You are the QA Reviewer (fast variant, open-weights edition). Your job is to verify that the diff matches the plan **semantically**, detect **scope creep**, and detect **plan drift**.
+
+Do not ask the user questions. Return `[PASS]` or `[FAIL]` only. If you're tempted to ask, FAIL instead and let the build agent fix it.
+
+**Always re-run tests, lint, and typecheck.** Do not skip verification steps. Run every command yourself before returning `[PASS]`.
+
+# Process
+
+1. **Read the plan** at the path provided.
+2. **Inspect the diff.** Run `git diff` (against merge base — try `git merge-base HEAD origin/main` then `origin/master`) and `git diff --stat`. Also run `git status` to see untracked files.
+3. **Plan-drift check (AUTO-FAIL).** For each modified file in the diff, verify it appears in the plan's `## File-level changes`. A modified file NOT listed in `## File-level changes` is AUTO-FAIL. Report as `Plan drift: <path> modified but not in ## File-level changes`.
+4. **Scope-creep check.** For each UNTRACKED file (from `git status`) that is NOT in `## File-level changes`, run `git log --oneline -- <file>` to determine whether the file is pre-existing work or scope creep. If the file has no prior commits on this branch AND isn't in the plan, FAIL with `Scope creep: <path> untracked and not in plan`.
+5. **Semantic verification.** For each item in `## File-level changes`, verify the corresponding code change exists and matches the description by reading the code. For each `## Acceptance criteria` item, verify it is actually met — do NOT trust `[x]` checkboxes.
+6. **Plan-state verify commands.** Run `bunx @glrs-dev/harness-plugin-opencode plan-check --run <plan-path>` to get the list of verify commands for pending items. Execute each one via `bash`. Any non-zero exit → FAIL with `Verify failed: <command> (exit N)`. If the plan has no fence (legacy), plan-check emits `legacy (no plan-state fence)` — skip this step.
+7. **Full-suite re-run.** Run the project's test / lint / typecheck commands (discover from `package.json` scripts / `Makefile` / `AGENTS.md`). Any failure → FAIL.
+8. **Scan for new tech debt.** Run `todo_scan` with `onlyChanged: true`. For every TODO / FIXME / HACK / XXX in the result, check whether the plan's `## Out of scope` or `## Open questions` section acknowledges it. Unacknowledged new debt → FAIL with the specific `file:line`.
+9. **AGENTS.md freshness (light check).** If the change shifts a convention documented in a local `AGENTS.md` in a touched directory, FAIL with `Update <path>/AGENTS.md to reflect <specific change>`.
+
+# Output
+
+Exactly one of these two formats. Nothing else.
+
+**If everything passes:**
+
+```
+[PASS]
+
+<2–3 sentence summary of verified changes.>
+```
+
+**If anything fails:**
+
+```
+[FAIL]
+
+1. <File:line> — <Specific issue>
+2. <File:line> — <Next issue>
+...
+```
+
+# Rules
+
+- Never suggest fixes. Report precisely; the build agent will fix.
+- Never trust the build agent's narrative. "Pre-existing work" requires `git log --oneline -- <file>` evidence.
+- A single failing item is enough to FAIL. Do not minimize.
+- **AUTO-FAIL on plan drift.** Modified file not in `## File-level changes` → FAIL, no exceptions.
+- **AUTO-FAIL on scope creep.** Untracked file not in plan with no prior commits → FAIL.
+- If the diff is large (>10 files or >500 lines) or touches high-risk paths (auth / crypto / billing / migrations), tell the PRIME to delegate to `@qa-thorough` instead.

package/dist/vendor/harness-opencode/dist/{chunk-WBBN7OVN.js → chunk-BWERBERN.js}

@@ -257,7 +257,7 @@ async function requirePlugin() {
     );
     process.exit(1);
   }
-  const { install: install2 } = await import("./install-X5KEANRB.js");
+  const { install: install2 } = await import("./install-5JKWK6Z4.js");
   await install2({ nonInteractive: true });
 }
 
@@ -764,6 +764,25 @@ ${c.bold}Ready.${c.reset} Run ${c.green}opencode${c.reset} to start.
         fast: [preset.fast]
       };
       ok(`Models configured`);
+      const midExecIdx = await promptChoice(
+        "  Use a strict executor for build agents? (recommended for Kimi/Qwen/DeepSeek)",
+        ["No (use mid model as reasoning builder)", "Yes (configure mid-execute model)"],
+        0
+      );
+      if (midExecIdx === 1) {
+        const { input } = await import("@inquirer/prompts");
+        const midExecModel = await input({
+          message: "  mid-execute model ID:",
+          default: preset.mid
+        });
+        if (midExecModel) {
+          pluginOpts.models["mid-execute"] = [midExecModel];
+          newModelsValue["mid-execute"] = [midExecModel];
+          info(`  mid-execute \u2192 ${midExecModel} (strict executor prompts)`);
+        }
+      } else {
+        info(`  mid-execute: skipped (build agents use mid model with reasoning prompts)`);
+      }
     } else if (!pluginOpts._skipModels) {
       info("Enter model IDs in <provider>/<model-id> format (e.g. amazon-bedrock/global.anthropic.claude-opus-4-7)");
       const { input } = await import("@inquirer/prompts");
@@ -771,17 +790,26 @@ ${c.bold}Ready.${c.reset} Run ${c.green}opencode${c.reset} to start.
       const midModel = await input({ message: "  mid (balanced):" });
       const fastModel = await input({ message: "  fast (cheapest):" });
       if (deepModel) {
+        const resolvedMid = midModel || deepModel;
         pluginOpts.models = {
           deep: [deepModel],
-          mid: [midModel || deepModel],
+          mid: [resolvedMid],
           fast: [fastModel || midModel || deepModel]
         };
         newModelsValue = {
           deep: [deepModel],
-          mid: [midModel || deepModel],
+          mid: [resolvedMid],
           fast: [fastModel || midModel || deepModel]
         };
         ok("Models: custom");
+        const midExecModel = await input({ message: "  mid-execute (optional strict executor, press Enter to skip):" });
+        if (midExecModel) {
+          pluginOpts.models["mid-execute"] = [midExecModel];
+          newModelsValue["mid-execute"] = [midExecModel];
+          info(`  mid-execute \u2192 ${midExecModel} (strict executor prompts)`);
+        } else {
+          info(`  mid-execute: skipped (build agents use mid model with reasoning prompts)`);
+        }
       } else {
         ok("Models: OpenCode defaults");
       }

package/dist/vendor/harness-opencode/dist/{chunk-CZMAJISX.js → chunk-EK7K4NTV.js}

@@ -47,7 +47,9 @@ function readPrompt(name) {
 var primePrompt = readPrompt("prime.md");
 var planPrompt = readPrompt("plan.md");
 var buildPrompt = readPrompt("build.md");
+var buildOpenPrompt = readPrompt("build.open.md");
 var qaReviewerPrompt = readPrompt("qa-reviewer.md");
+var qaReviewerOpenPrompt = readPrompt("qa-reviewer.open.md");
 var qaThoroughPrompt = readPrompt("qa-thorough.md");
 var planReviewerPrompt = readPrompt("plan-reviewer.md");
 var codeSearcherPrompt = readPrompt("code-searcher.md");
@@ -57,11 +59,24 @@ var docsMaintainerPrompt = readPrompt("docs-maintainer.md");
 var libReaderPrompt = readPrompt("lib-reader.md");
 var agentsMdWriterPrompt = readPrompt("agents-md-writer.md");
 var pilotBuilderPrompt = readPrompt("pilot-builder.md");
+var pilotBuilderOpenPrompt = readPrompt("pilot-builder.open.md");
 var pilotPlannerPrompt = readPrompt("pilot-planner.md");
 var researchPrompt = readPrompt("research.md");
 var researchWebPrompt = readPrompt("research-web.md");
 var researchLocalPrompt = readPrompt("research-local.md");
 var researchAutoPrompt = readPrompt("research-auto.md");
+var EXECUTOR_VARIANT_AGENTS = {
+  build: { reasoning: buildPrompt, strict: buildOpenPrompt },
+  "qa-reviewer": { reasoning: qaReviewerPrompt, strict: qaReviewerOpenPrompt },
+  "pilot-builder": { reasoning: pilotBuilderPrompt, strict: pilotBuilderOpenPrompt }
+};
+function getStrictPrompt(agentName) {
+  const variants = EXECUTOR_VARIANT_AGENTS[agentName];
+  if (!variants) {
+    throw new Error(`getStrictPrompt: no strict variant registered for agent "${agentName}"`);
+  }
+  return variants.strict;
+}
 function stripFrontmatter(md) {
   if (!md.startsWith("---")) return md;
   const end = md.indexOf("\n---", 3);
@@ -563,12 +578,12 @@ var AGENT_TIERS = {
   "research-web": "deep",
   "research-local": "deep",
   "research-auto": "deep",
-  build: "mid",
-  "qa-reviewer": "mid",
+  build: "mid-execute",
+  "qa-reviewer": "mid-execute",
+  "pilot-builder": "mid-execute",
   "docs-maintainer": "mid",
   "lib-reader": "mid",
   "agents-md-writer": "mid",
-  "pilot-builder": "mid",
   "code-searcher": "fast"
 };
 function createAgents() {
@@ -724,6 +739,7 @@ function formatModelOverrideWarning(id, source, suggestion) {
 }
 
 export {
+  getStrictPrompt,
   AGENT_TIERS,
   createAgents,
   validateModelOverride,

package/dist/vendor/harness-opencode/dist/cli.js

@@ -2,7 +2,7 @@
 import {
   createAgents,
   validateModelOverride
-} from "./chunk-CZMAJISX.js";
+} from "./chunk-EK7K4NTV.js";
 import {
   getSessionsPath,
   registerSession,
@@ -11,7 +11,7 @@ import {
 import {
   install,
   requirePlugin
-} from "./chunk-WBBN7OVN.js";
+} from "./chunk-BWERBERN.js";
 import "./chunk-VJUETC6A.js";
 import {
   getPilotDir,
@@ -1142,11 +1142,60 @@ CREATE TABLE IF NOT EXISTS events (
 CREATE INDEX IF NOT EXISTS idx_events_run ON events(run_id, id);
 CREATE INDEX IF NOT EXISTS idx_events_run_task ON events(run_id, task_id, id);
 `.trim();
+var V2_SQL = `
+CREATE TABLE IF NOT EXISTS workflows (
+  id            TEXT    NOT NULL PRIMARY KEY,
+  goal          TEXT    NOT NULL,
+  started_at    INTEGER NOT NULL,
+  finished_at   INTEGER,
+  status        TEXT    NOT NULL CHECK (status IN ('pending','running','completed','aborted','failed')),
+  current_phase TEXT
+);
+
+CREATE TABLE IF NOT EXISTS phases (
+  workflow_id   TEXT    NOT NULL,
+  name          TEXT    NOT NULL CHECK (name IN ('scope','plan','build','qa','followup')),
+  status        TEXT    NOT NULL CHECK (status IN ('pending','running','completed','aborted','failed')),
+  started_at    INTEGER,
+  finished_at   INTEGER,
+  artifact_path TEXT,
+  PRIMARY KEY (workflow_id, name),
+  FOREIGN KEY (workflow_id) REFERENCES workflows(id) ON DELETE CASCADE
+);
+
+CREATE TABLE IF NOT EXISTS artifacts (
+  id          INTEGER PRIMARY KEY AUTOINCREMENT,
+  workflow_id TEXT    NOT NULL,
+  phase       TEXT    NOT NULL,
+  kind        TEXT    NOT NULL,
+  path        TEXT    NOT NULL,
+  created_at  INTEGER NOT NULL,
+  sha256      TEXT,
+  FOREIGN KEY (workflow_id) REFERENCES workflows(id) ON DELETE CASCADE
+);
+
+CREATE INDEX IF NOT EXISTS idx_artifacts_workflow_phase ON artifacts(workflow_id, phase);
+
+ALTER TABLE events ADD COLUMN phase TEXT;
+
+INSERT INTO workflows (id, goal, started_at, finished_at, status, current_phase)
+SELECT id, plan_slug, started_at, finished_at, status, 'build' FROM runs;
+
+INSERT INTO phases (workflow_id, name, status, started_at, finished_at, artifact_path)
+SELECT id, 'build', status, started_at, finished_at, NULL FROM runs;
+
+UPDATE events SET phase = 'build' WHERE phase IS NULL;
+`.trim();
 var MIGRATIONS = [
   {
     version: 1,
     description: "initial pilot schema (runs/tasks/events)",
     sql: V1_SQL
+  },
+  {
+    version: 2,
+    description: "workflows/phases/artifacts tables + events.phase column",
+    sql: V2_SQL
   }
 ];
 function applyMigrations(db) {
@@ -1279,8 +1328,8 @@ function appendEvent(db, args) {
     });
   }
   db.run(
-    `INSERT INTO events (run_id, task_id, ts, kind, payload) VALUES (?, ?, ?, ?, ?)`,
-    [args.runId, args.taskId ?? null, ts, args.kind, payloadStr]
+    `INSERT INTO events (run_id, task_id, ts, kind, payload, phase) VALUES (?, ?, ?, ?, ?, ?)`,
+    [args.runId, args.taskId ?? null, ts, args.kind, payloadStr, args.phase ?? null]
   );
   if (eventSubscribers.length > 0) {
     const snapshot = eventSubscribers.slice();
@@ -1291,6 +1340,7 @@ function appendEvent(db, args) {
           taskId: args.taskId ?? null,
           kind: args.kind,
           payload: args.payload,
+          phase: args.phase ?? null,
           ts
         });
       } catch {
@@ -1865,25 +1915,78 @@ function fixPrompt(_task, last) {
   return sections.join("\n");
 }
 
-// src/pilot/verify/runner.ts
-import { spawn as spawn2 } from "child_process";
-var DEFAULT_TIMEOUT_MS = 5 * 60 * 1e3;
-var DEFAULT_OUTPUT_CAP_BYTES = 256 * 1024;
-var TRUNCATION_NOTICE = "\n[pilot] verify output truncated\n";
-async function runVerify(commands, options) {
+// src/pilot/gates/composite.ts
+async function evalAllGate(gate, ctx) {
+  const startedAt = Date.now();
   const results = [];
-  for (const command10 of commands) {
-    const result = await runOne(command10, options);
-    results.push(result);
-    if (!result.ok) {
-      return { ok: false, results, failure: result };
+  for (const sub of gate.gates) {
+    const subResult = await evalGate(sub, ctx);
+    results.push({ gate: sub, result: subResult });
+    if (!subResult.ok) {
+      const evidence2 = {
+        kind: "all",
+        results,
+        failure: subResult
+      };
+      return {
+        ok: false,
+        reason: subResult.reason,
+        evidence: evidence2,
+        durationMs: Date.now() - startedAt
+      };
     }
   }
+  const evidence = { kind: "all", results };
   return {
     ok: true,
-    results
+    evidence,
+    durationMs: Date.now() - startedAt
   };
 }
+async function evalAnyGate(gate, ctx) {
+  const startedAt = Date.now();
+  const results = [];
+  if (gate.gates.length === 0) {
+    const evidence2 = { kind: "any", results };
+    return {
+      ok: false,
+      reason: "any-gate has no sub-gates to satisfy",
+      evidence: evidence2,
+      durationMs: Date.now() - startedAt
+    };
+  }
+  let lastResult = null;
+  for (const sub of gate.gates) {
+    const subResult = await evalGate(sub, ctx);
+    results.push({ gate: sub, result: subResult });
+    lastResult = subResult;
+    if (subResult.ok) {
+      const evidence2 = { kind: "any", results };
+      return {
+        ok: true,
+        evidence: evidence2,
+        durationMs: Date.now() - startedAt
+      };
+    }
+  }
+  const evidence = {
+    kind: "any",
+    results,
+    failure: lastResult ?? void 0
+  };
+  return {
+    ok: false,
+    reason: `any-gate exhausted: all ${results.length} sub-gates failed`,
+    evidence,
+    durationMs: Date.now() - startedAt
+  };
+}
+
+// src/pilot/verify/spawn.ts
+import { spawn as spawn2 } from "child_process";
+var DEFAULT_TIMEOUT_MS = 5 * 60 * 1e3;
+var DEFAULT_OUTPUT_CAP_BYTES = 256 * 1024;
+var TRUNCATION_NOTICE = "\n[pilot] verify output truncated\n";
 async function runOne(command10, options) {
   if (typeof command10 !== "string" || command10.length === 0) {
     throw new TypeError(`runOne: command must be a non-empty string`);
@@ -2020,6 +2123,147 @@ function killTree(child) {
   }, 2e3).unref();
 }
 
+// src/pilot/gates/shell.ts
+async function evalShellGate(gate, ctx) {
+  const result = await runOne(gate.command, {
+    cwd: ctx.cwd,
+    env: ctx.env,
+    abortSignal: ctx.abortSignal,
+    onLine: ctx.onShellLine,
+    timeoutMs: gate.timeoutMs,
+    outputCapBytes: ctx.shellOutputCapBytes
+  });
+  return toGateResult(result);
+}
+function toGateResult(result) {
+  if (result.ok) {
+    return {
+      ok: true,
+      durationMs: result.durationMs,
+      evidence: { kind: "shell", result }
+    };
+  }
+  const reason = formatShellFailure(result);
+  return {
+    ok: false,
+    reason,
+    durationMs: result.durationMs,
+    evidence: { kind: "shell", result }
+  };
+}
+function formatShellFailure(result) {
+  const flags = [];
+  if (result.timedOut) flags.push("timed-out");
+  if (result.aborted) flags.push("aborted");
+  if (result.signal) flags.push(`signal=${result.signal}`);
+  const flagSuffix = flags.length > 0 ? ` [${flags.join(",")}]` : "";
+  return `shell gate failed: ${result.command} \u2192 exit ${result.exitCode}${flagSuffix}`;
+}
+
+// src/pilot/gates/eval.ts
+async function evalGate(gate, ctx) {
+  switch (gate.kind) {
+    case "shell":
+      return evalShellGate(gate, ctx);
+    case "all":
+      return evalAllGate(gate, ctx);
+    case "any":
+      return evalAnyGate(gate, ctx);
+    default: {
+      const _exhaustive = gate;
+      throw new Error(
+        `evalGate: unknown gate kind ${_exhaustive.kind}`
+      );
+    }
+  }
+}
+
+// src/pilot/gates/types.ts
+function asShellEvidence(evidence) {
+  if (typeof evidence === "object" && evidence !== null && evidence.kind === "shell") {
+    return evidence;
+  }
+  return null;
+}
+function asCompositeEvidence(evidence) {
+  if (typeof evidence === "object" && evidence !== null && (evidence.kind === "all" || evidence.kind === "any")) {
+    return evidence;
+  }
+  return null;
+}
+
+// src/pilot/verify/runner.ts
+async function runVerify(commands, options) {
+  if (commands.length === 0) {
+    return { ok: true, results: [] };
+  }
+  const gate = {
+    kind: "all",
+    gates: commands.map((command10) => ({
+      kind: "shell",
+      command: command10,
+      timeoutMs: options.timeoutMs
+    }))
+  };
+  const ctx = {
+    cwd: options.cwd,
+    env: options.env,
+    abortSignal: options.abortSignal,
+    onShellLine: options.onLine,
+    shellOutputCapBytes: options.outputCapBytes
+  };
+  const gateResult = await evalGate(gate, ctx);
+  return toRunVerifyResult(gateResult);
+}
+function toRunVerifyResult(gateResult) {
+  const composite = asCompositeEvidence(gateResult.evidence);
+  if (composite === null || composite.kind !== "all") {
+    throw new Error(
+      `runVerify: expected composite all-gate evidence, got ${gateResultDescriptor(gateResult)}`
+    );
+  }
+  const results = composite.results.map((entry) => extractCommandResult(entry));
+  if (gateResult.ok) {
+    return {
+      ok: true,
+      results
+    };
+  }
+  const failingEntry = composite.results[composite.results.length - 1];
+  if (!failingEntry || failingEntry.result.ok) {
+    throw new Error(
+      "runVerify: all-gate failed but no failing sub-result was recorded"
+    );
+  }
+  const failureCommandResult = extractCommandResult(failingEntry);
+  if (failureCommandResult.ok) {
+    throw new Error(
+      "runVerify: failing sub-gate produced a successful CommandResult"
+    );
+  }
+  return {
+    ok: false,
+    results,
+    failure: failureCommandResult
+  };
+}
+function extractCommandResult(entry) {
+  const shell = asShellEvidence(entry.result.evidence);
+  if (shell === null) {
+    throw new Error(
+      `runVerify: expected shell-gate evidence in all-gate child, got ${gateResultDescriptor(entry.result)}`
+    );
+  }
+  return shell.result;
+}
+function gateResultDescriptor(result) {
+  const evidence = result.evidence;
+  return JSON.stringify({
+    ok: result.ok,
+    evidenceKind: evidence?.kind ?? null
+  });
+}
+
 // src/pilot/verify/touches.ts
 import picomatch2 from "picomatch";
 import { execFile as execFile2 } from "child_process";
@@ -2530,7 +2774,11 @@ async function runOneTaskImpl(deps, task, opts) {
           command: f.command,
           exitCode: f.exitCode,
           output: f.output.slice(0, 4096),
-          reason: reason2
+          reason: reason2,
+          // Step 1 of pilot redesign: gate descriptor on every
+          // verify-derived event. Future LLM/approval gates emit
+          // identically-shaped events with a different `gate.kind`.
+          gate: { kind: "shell", command: f.command }
         }
       });
       return;
@@ -2539,7 +2787,10 @@ async function runOneTaskImpl(deps, task, opts) {
       runId: deps.runId,
       taskId: task.id,
       kind: "task.baseline.passed",
-      payload: { commands: allVerify.length }
+      payload: {
+        commands: allVerify.length,
+        gate: { kind: "all", subKind: "shell", count: baselineVerify.length }
+      }
     });
   }
   let lastFailure = null;
@@ -2695,7 +2946,8 @@ async function runOneTaskImpl(deps, task, opts) {
           exitCode: lastFailure.exitCode,
           timedOut: verifyResult.failure.timedOut,
           aborted: verifyResult.failure.aborted,
-          output: verifyResult.failure.output.slice(-2048)
+          output: verifyResult.failure.output.slice(-2048),
+          gate: { kind: "shell", command: lastFailure.command }
         }
       });
       if (verifyResult.failure.aborted) {
@@ -2721,7 +2973,10 @@ async function runOneTaskImpl(deps, task, opts) {
       runId: deps.runId,
       taskId: task.id,
       kind: "task.verify.passed",
-      payload: { attempt }
+      payload: {
+        attempt,
+        gate: { kind: "all", subKind: "shell", count: allVerify.length }
+      }
     });
     const touches = await enforceTouches({
       cwd,
@@ -3311,7 +3566,7 @@ function startStreamingLogger(args) {
   const taskStart = /* @__PURE__ */ new Map();
   let succeeded = 0;
   let failed = 0;
-  const INLINE_BLOCKED_CAP = 5;
+  const INLINE_BLOCKED_CAP = 0;
   let blockedCount = 0;
   let blockedInlineEmitted = 0;
   let blockedOverflowEmitted = false;
@@ -3350,6 +3605,24 @@ function startStreamingLogger(args) {
         if (id !== null) taskStart.set(id, event.ts);
         write(`task.started ${id ?? "?"}`);
         break;
+      case "task.baseline.passed":
+        break;
+      case "task.baseline.failed": {
+        const bp = event.payload;
+        if (bp !== null && typeof bp === "object" && typeof bp.command === "string" && typeof bp.exitCode === "number") {
+          write(
+            `task.baseline.failed ${id ?? "?"} (${bp.command} \u2192 exit ${bp.exitCode})`
+          );
+          const output = typeof bp.output === "string" ? bp.output : null;
+          if (output !== null && output.trim().length > 0) {
+            const tail = output.trim().split("\n").slice(-6).map((l) => `    ${l}`).join("\n");
+            writeRaw(tail);
+          }
+        } else {
+          write(`task.baseline.failed ${id ?? "?"}`);
+        }
+        break;
+      }
       case "task.verify.passed":
         write(`task.verify.passed ${id ?? "?"}`);
         break;
@@ -3435,7 +3708,7 @@ function startStreamingLogger(args) {
       case "task.attempt": {
         const p = event.payload;
         if (p !== null && typeof p === "object" && typeof p.attempt === "number" && typeof p.of === "number" && p.attempt >= 2) {
-          writeRaw(`  attempt ${p.attempt}/${p.of} (retry with fix prompt)`);
+          write(`task.retry ${id ?? "?"} attempt ${p.attempt}/${p.of}`);
         }
         break;
       }
@@ -3561,9 +3834,17 @@ Failed tasks (${failed.length}):
     session:  ${session}
     worktree: ${worktree}
     elapsed:  ${elapsed}   attempts: ${t.attempts}
-
 `
         );
+        const baselineOutput = resolveBaselineOutput(db, runId, t.task_id);
+        if (baselineOutput !== null) {
+          const tail = baselineOutput.trim().split("\n").slice(-6).map((l) => `      ${l}`).join("\n");
+          process.stdout.write(`    output:
+${tail}
+`);
+        }
+        process.stdout.write(`
+`);
       }
     }
   }
@@ -3592,6 +3873,18 @@ function resolveFailureDetail(db, runId, row) {
     reason: row.last_error ?? "(no reason recorded)"
   };
 }
+function resolveBaselineOutput(db, runId, taskId) {
+  const events = readEventsDecoded(db, { runId, taskId });
+  for (let i = events.length - 1; i >= 0; i--) {
+    const e = events[i];
+    if (e.kind !== "task.baseline.failed") continue;
+    const p = e.payload;
+    if (p !== null && typeof p === "object" && typeof p.output === "string") {
+      return p.output;
+    }
+  }
+  return null;
+}
 function truncateSummary(s, maxChars) {
   if (s.length <= maxChars) return s;
   return s.slice(0, maxChars - 1) + "\u2026";