@tea-agent/loop-agent 0.35.1-beta.0 → 0.35.1-beta.2

package/skills/verification-before-completion/SKILL.md

@@ -1,154 +1,154 @@
----
-name: verification-before-completion
-description: 在宣称 work complete、fixed 或 passing，或在 commit / 创建 PR 之前使用——须先运行 verification commands 并确认 output，再作任何 success claims；始终 evidence before assertions
----
-
-# Verification Before Completion
-
-## Overview
-
-未经验证就宣称 work complete 是不诚实，不是效率。
-
-**Core principle：** 始终 evidence before claims。
-
-**违反本条字面即违反其精神。**
-
-## The Iron Law
-
-```
-NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
-```
-
-若本 message 中尚未运行 verification command，不得宣称 passes。
-
-## The Gate Function
-
-```
-BEFORE claiming any status or expressing satisfaction:
-
-1. IDENTIFY: What command proves this claim?
-2. RUN: Execute the FULL command (fresh, complete)
-3. READ: Full output, check exit code, count failures
-4. VERIFY: Does output confirm the claim?
-   - If NO: State actual status with evidence
-   - If YES: State claim WITH evidence
-5. ONLY THEN: Make the claim
-
-Skip any step = lying, not verifying
-```
-
-## Common Failures
-
-| Claim | Requires | Not Sufficient |
-|-------|----------|----------------|
-| Tests pass | Test command output: 0 failures | Previous run, "should pass" |
-| Linter clean | Linter output: 0 errors | Partial check, extrapolation |
-| Build succeeds | Build command: exit 0 | Linter passing, logs look good |
-| Bug fixed | Test original symptom: passes | Code changed, assumed fixed |
-| Regression test works | Red-green cycle verified | Test passes once |
-| Agent completed | VCS diff shows changes | Agent reports "success" |
-| Requirements met | Line-by-line checklist | Tests passing |
-| Harness integrity | `scripts/check-repo.sh` exit 0 | Files look correct |
-| Harness CI | `scripts/ci.sh` exit 0 | Individual checks pass |
-| Contract handoff | Handoff checklist completed + progress/report updated | "Should be fine"
-
-## Red Flags - STOP
-
-- 使用 "should"、"probably"、"seems to"
-- 验证前表达满意（"Great!"、"Perfect!"、"Done!" 等）
-- 未验证就要 commit/push/PR
-- 信任 agent success reports
-- 依赖 partial verification
-- 认为 "just this once"
-- 疲惫想结束工作
-- **任何未运行 verification 却暗示 success 的措辞**
-
-## Rationalization Prevention
-
-| Excuse | Reality |
-|--------|---------|
-| "Should work now" | RUN the verification |
-| "I'm confident" | Confidence ≠ evidence |
-| "Just this once" | No exceptions |
-| "Linter passed" | Linter ≠ compiler |
-| "Agent said success" | Verify independently |
-| "I'm tired" | Exhaustion ≠ excuse |
-| "Partial check is enough" | Partial proves nothing |
-| "Different words so rule doesn't apply" | Spirit over letter |
-
-## Key Patterns
-
-**Tests:**
-```
-✅ [Run test command] [See: 34/34 pass] "All tests pass"
-❌ "Should pass now" / "Looks correct"
-```
-
-**Regression tests (TDD Red-Green):**
-```
-✅ Write → Run (pass) → Revert fix → Run (MUST FAIL) → Restore → Run (pass)
-❌ "I've written a regression test" (without red-green verification)
-```
-
-**Build:**
-```
-✅ [Run build] [See: exit 0] "Build passes"
-❌ "Linter passed" (linter doesn't check compilation)
-```
-
-**Requirements:**
-```
-✅ Re-read plan → Create checklist → Verify each → Report gaps or completion
-❌ "Tests pass, phase complete"
-```
-
-**Agent delegation:**
-```
-✅ Agent reports success → Check VCS diff → Verify changes → Report actual state
-❌ Trust agent report
-```
-
-## Harness-Specific Verification
-
-在 harness-governed repo 中工作（存在 `harness.json`）时：
-
-- **Docs/structure changes** → `bash scripts/check-repo.sh`
-- **Full-repo delivery** → `bash scripts/ci.sh`
-- **Cross-platform changes** → 验证 OpenCode 与 Pi-Agent 两条路径
-- **Contract changes** → 验证 contract docs 已更新 + tests 对齐
-- **Handoff** → 宣称 complete 前运行 `handoff check`
-
-完整 command 选择见项目 `ai_workspace/loop-agent/verification-matrix.md`。
-
-## Why This Matters
-
-来自 24 条 failure memories：
-- human partner 说 "I don't believe you" — trust 已破裂
-- Undefined functions 已 ship — 会 crash
-- Missing requirements 已 ship — 功能不完整
-- 虚假完成浪费时间 → redirect → rework
-- 违反："Honesty is a core value. If you lie, you'll be replaced."
-
-## When To Apply
-
-**在以下情况之前 ALWAYS：**
-- 任何 success/completion claims 的变体
-- 任何表达满意
-- 任何关于 work state 的正面陈述
-- Commit、PR creation、task completion
-- 进入 next task
-- 委派给 agents
-
-**规则适用于：**
-- 精确短语
--  paraphrases 与同义词
-- success 的暗示
-- 任何暗示 completion/correctness 的沟通
-
-## The Bottom Line
-
-**Verification 无捷径。**
-
-Run the command. Read the output. THEN claim the result.
-
-This is non-negotiable.
+---
+name: verification-before-completion
+description: 在宣称 work complete、fixed 或 passing，或在 commit / 创建 PR 之前使用——须先运行 verification commands 并确认 output，再作任何 success claims；始终 evidence before assertions
+---
+
+# Verification Before Completion
+
+## Overview
+
+未经验证就宣称 work complete 是不诚实，不是效率。
+
+**Core principle：** 始终 evidence before claims。
+
+**违反本条字面即违反其精神。**
+
+## The Iron Law
+
+```
+NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
+```
+
+若本 message 中尚未运行 verification command，不得宣称 passes。
+
+## The Gate Function
+
+```
+BEFORE claiming any status or expressing satisfaction:
+
+1. IDENTIFY: What command proves this claim?
+2. RUN: Execute the FULL command (fresh, complete)
+3. READ: Full output, check exit code, count failures
+4. VERIFY: Does output confirm the claim?
+   - If NO: State actual status with evidence
+   - If YES: State claim WITH evidence
+5. ONLY THEN: Make the claim
+
+Skip any step = lying, not verifying
+```
+
+## Common Failures
+
+| Claim | Requires | Not Sufficient |
+|-------|----------|----------------|
+| Tests pass | Test command output: 0 failures | Previous run, "should pass" |
+| Linter clean | Linter output: 0 errors | Partial check, extrapolation |
+| Build succeeds | Build command: exit 0 | Linter passing, logs look good |
+| Bug fixed | Test original symptom: passes | Code changed, assumed fixed |
+| Regression test works | Red-green cycle verified | Test passes once |
+| Agent completed | VCS diff shows changes | Agent reports "success" |
+| Requirements met | Line-by-line checklist | Tests passing |
+| Harness integrity | `scripts/check-repo.sh` exit 0 | Files look correct |
+| Harness CI | `scripts/ci.sh` exit 0 | Individual checks pass |
+| Contract handoff | Handoff checklist completed + progress/report updated | "Should be fine"
+
+## Red Flags - STOP
+
+- 使用 "should"、"probably"、"seems to"
+- 验证前表达满意（"Great!"、"Perfect!"、"Done!" 等）
+- 未验证就要 commit/push/PR
+- 信任 agent success reports
+- 依赖 partial verification
+- 认为 "just this once"
+- 疲惫想结束工作
+- **任何未运行 verification 却暗示 success 的措辞**
+
+## Rationalization Prevention
+
+| Excuse | Reality |
+|--------|---------|
+| "Should work now" | RUN the verification |
+| "I'm confident" | Confidence ≠ evidence |
+| "Just this once" | No exceptions |
+| "Linter passed" | Linter ≠ compiler |
+| "Agent said success" | Verify independently |
+| "I'm tired" | Exhaustion ≠ excuse |
+| "Partial check is enough" | Partial proves nothing |
+| "Different words so rule doesn't apply" | Spirit over letter |
+
+## Key Patterns
+
+**Tests:**
+```
+✅ [Run test command] [See: 34/34 pass] "All tests pass"
+❌ "Should pass now" / "Looks correct"
+```
+
+**Regression tests (TDD Red-Green):**
+```
+✅ Write → Run (pass) → Revert fix → Run (MUST FAIL) → Restore → Run (pass)
+❌ "I've written a regression test" (without red-green verification)
+```
+
+**Build:**
+```
+✅ [Run build] [See: exit 0] "Build passes"
+❌ "Linter passed" (linter doesn't check compilation)
+```
+
+**Requirements:**
+```
+✅ Re-read plan → Create checklist → Verify each → Report gaps or completion
+❌ "Tests pass, phase complete"
+```
+
+**Agent delegation:**
+```
+✅ Agent reports success → Check VCS diff → Verify changes → Report actual state
+❌ Trust agent report
+```
+
+## Harness-Specific Verification
+
+在 harness-governed repo 中工作（存在 `harness.json`）时：
+
+- **Docs/structure changes** → `bash scripts/check-repo.sh`
+- **Full-repo delivery** → `bash scripts/ci.sh`
+- **Cross-platform changes** → 验证 OpenCode 与 Pi-Agent 两条路径
+- **Contract changes** → 验证 contract docs 已更新 + tests 对齐
+- **Handoff** → 宣称 complete 前运行 `handoff check`
+
+完整 command 选择见项目 `ai_workspace/loop-agent/verification-matrix.md`。
+
+## Why This Matters
+
+来自 24 条 failure memories：
+- human partner 说 "I don't believe you" — trust 已破裂
+- Undefined functions 已 ship — 会 crash
+- Missing requirements 已 ship — 功能不完整
+- 虚假完成浪费时间 → redirect → rework
+- 违反："Honesty is a core value. If you lie, you'll be replaced."
+
+## When To Apply
+
+**在以下情况之前 ALWAYS：**
+- 任何 success/completion claims 的变体
+- 任何表达满意
+- 任何关于 work state 的正面陈述
+- Commit、PR creation、task completion
+- 进入 next task
+- 委派给 agents
+
+**规则适用于：**
+- 精确短语
+-  paraphrases 与同义词
+- success 的暗示
+- 任何暗示 completion/correctness 的沟通
+
+## The Bottom Line
+
+**Verification 无捷径。**
+
+Run the command. Read the output. THEN claim the result.
+
+This is non-negotiable.

package/skills/webapp-testing/SKILL.md

@@ -1,19 +1,19 @@
----
-name: webapp-testing
-description: 任务明确涉及 browser 渲染行为时，用于前端或本地 web UI 验证。
----
-
-# Webapp Testing
-
-仅当任务包含 browser UI 或本地 web app 时使用本 skill。
-
-## 规则
-
-- 优先使用项目现有的 dev server 与 test tooling。
-- 当 visual 或 interaction 行为重要时，用 browser 或文档化的 UI test 命令验证渲染行为。
-- UI 有变更时，检查 desktop 与 mobile 布局的 overlap、clipping、blank state。
-- 默认不添加 networked services 或第三方 scan。
-
-## Output
-
-报告确切的 server 命令、URL、browser/test 命令与观察结果。
+---
+name: webapp-testing
+description: 任务明确涉及 browser 渲染行为时，用于前端或本地 web UI 验证。
+---
+
+# Webapp Testing
+
+仅当任务包含 browser UI 或本地 web app 时使用本 skill。
+
+## 规则
+
+- 优先使用项目现有的 dev server 与 test tooling。
+- 当 visual 或 interaction 行为重要时，用 browser 或文档化的 UI test 命令验证渲染行为。
+- UI 有变更时，检查 desktop 与 mobile 布局的 overlap、clipping、blank state。
+- 默认不添加 networked services 或第三方 scan。
+
+## Output
+
+报告确切的 server 命令、URL、browser/test 命令与观察结果。

package/dist/worker/console/operation-wait.js

@@ -1,241 +0,0 @@
-import { projectOperationEventSummary, projectOperationForChat, } from "./chat/chat-event-store.js";
-import { isTerminalOperationState, } from "./operation-store.js";
-/**
- * P2: read-only event-driven long poll on the canonical operation event ring
- * (2026-08-13 Operator Chat long-run supervision). Server contract bounds:
- * the model-facing `maxWaitMs` is clamped to [minWaitMs, maxWaitMsBound];
- * defaults allow the 60–180s model supervision cadence and tests may inject
- * short bounds.
- */
-export const DEFAULT_OPERATION_WAIT_MIN_MS = 60_000;
-export const DEFAULT_OPERATION_WAIT_MAX_MS = 180_000;
-export class OperationWaitError extends Error {
-    code;
-    constructor(code, message) {
-        super(message);
-        this.code = code;
-        this.name = "OperationWaitError";
-    }
-}
-function isFocusedOperation(operation) {
-    return (isTerminalOperationState(operation.state) ||
-        operation.state === "needs-reconcile");
-}
-function settledSummary(input) {
-    const projectedEvents = input.newEvents.length > 0
-        ? input.newEvents.map(projectOperationEventSummary)
-        : [];
-    const nextSeq = input.newEvents.length > 0
-        ? input.newEvents[input.newEvents.length - 1].seq
-        : input.afterSeq;
-    return {
-        operationId: input.operationId,
-        state: input.operation.state,
-        changed: input.changed,
-        timedOut: input.timedOut,
-        events: projectedEvents,
-        nextSeq,
-        operation: projectOperationForChat(input.operation),
-    };
-}
-/**
- * Deterministic read-only wait over the canonical operation event stream
- * (AC-003 / AC-004). Completion paths:
- *  - existing events (seq > afterSeq) → immediate (changed: true);
- *  - operation terminal/needs-reconcile with unconsumed events → immediate summary (changed: true);
- *  - operation terminal/needs-reconcile with no new events → immediate summary (changed: false);
- *  - EVENT_CURSOR_EXPIRED when the cursor fell out of the retained ring;
- *  - first subscribed event/state change → immediate settle;
- *  - maxWaitMs elapsed with no change → timedOut: true summary (not a failure).
- *
- * Race safety: the listener is registered BEFORE listFrom, closing the
- * listFrom/subscribe gap; every path settles exactly once through a guarded
- * `finish`, and listener + timer are always cleaned up on settle.
- */
-export async function waitForOperationChange(input) {
-    const operationId = input.operationId?.trim();
-    if (!operationId) {
-        throw new OperationWaitError("INVALID_INPUT", "operationId is required");
-    }
-    if (!Number.isInteger(input.afterSeq) || input.afterSeq < 0) {
-        throw new OperationWaitError("INVALID_INPUT", "afterSeq must be a non-negative integer");
-    }
-    const minWaitMs = input.minWaitMs ?? DEFAULT_OPERATION_WAIT_MIN_MS;
-    const maxWaitMsBound = input.maxWaitMsBound ?? DEFAULT_OPERATION_WAIT_MAX_MS;
-    if (!Number.isFinite(minWaitMs) ||
-        !Number.isFinite(maxWaitMsBound) ||
-        minWaitMs < 0 ||
-        maxWaitMsBound < minWaitMs) {
-        throw new OperationWaitError("INVALID_INPUT", "invalid wait bounds");
-    }
-    const rawMaxWaitMs = input.maxWaitMs ?? maxWaitMsBound;
-    if (!Number.isFinite(rawMaxWaitMs) || rawMaxWaitMs < 0) {
-        throw new OperationWaitError("INVALID_INPUT", "maxWaitMs must be a non-negative number");
-    }
-    const maxWaitMs = Math.max(minWaitMs, Math.min(maxWaitMsBound, Math.floor(rawMaxWaitMs)));
-    const afterSeq = input.afterSeq;
-    const operation = await input.operations.get(operationId);
-    if (!operation) {
-        throw new OperationWaitError("NOT_FOUND", `operation not found: ${operationId}`);
-    }
-    const events = input.events;
-    if (isFocusedOperation(operation)) {
-        // Flush unconsumed events before the terminal summary (AC-002): the
-        // caller's cursor must advance past every retained canonical event.
-        const listed = events.listFrom(operationId, afterSeq);
-        if ("error" in listed) {
-            throw new OperationWaitError("EVENT_CURSOR_EXPIRED", "event cursor expired; re-read operation snapshot");
-        }
-        if (listed.events.length > 0) {
-            return settledSummary({
-                operationId,
-                operation,
-                afterSeq,
-                changed: true,
-                timedOut: false,
-                newEvents: listed.events,
-            });
-        }
-        return settledSummary({
-            operationId,
-            operation,
-            afterSeq,
-            changed: false,
-            timedOut: false,
-            newEvents: [],
-        });
-    }
-    return new Promise((resolve, reject) => {
-        let settled = false;
-        let timer;
-        let unsubscribe;
-        const cleanup = () => {
-            if (timer !== undefined) {
-                clearTimeout(timer);
-                timer = undefined;
-            }
-            if (unsubscribe) {
-                unsubscribe();
-                unsubscribe = undefined;
-            }
-        };
-        const finish = (result) => {
-            if (settled)
-                return;
-            settled = true;
-            cleanup();
-            if (result instanceof OperationWaitError)
-                reject(result);
-            else
-                resolve(result);
-        };
-        const settleWithEvent = (event) => {
-            // Future-cursor guard (AC-003): events at or below the caller's
-            // afterSeq are already consumed and must not settle this wait.
-            if (event.seq <= afterSeq)
-                return;
-            // Listener path: async re-read of the operation for a fresh summary.
-            void (async () => {
-                try {
-                    const current = await input.operations.get(operationId);
-                    finish(settledSummary({
-                        operationId,
-                        operation: current ?? operation,
-                        afterSeq,
-                        changed: true,
-                        timedOut: false,
-                        newEvents: [event],
-                    }));
-                }
-                catch (error) {
-                    finish(error instanceof OperationWaitError
-                        ? error
-                        : new OperationWaitError("INVALID_INPUT", error instanceof Error ? error.message : String(error)));
-                }
-            })();
-        };
-        // Terminal recheck before subscribing: events landing during this await
-        // are still in the ring and are caught by listFrom below.
-        void (async () => {
-            try {
-                const current = await input.operations.get(operationId);
-                if (current && isFocusedOperation(current)) {
-                    // Same terminal flush as the initial path (AC-002): events
-                    // landing during the await are still in the retained ring.
-                    const listed = events.listFrom(operationId, afterSeq);
-                    if ("error" in listed) {
-                        finish(new OperationWaitError("EVENT_CURSOR_EXPIRED", "event cursor expired; re-read operation snapshot"));
-                        return;
-                    }
-                    if (listed.events.length > 0) {
-                        finish(settledSummary({
-                            operationId,
-                            operation: current,
-                            afterSeq,
-                            changed: true,
-                            timedOut: false,
-                            newEvents: listed.events,
-                        }));
-                        return;
-                    }
-                    finish(settledSummary({
-                        operationId,
-                        operation: current,
-                        afterSeq,
-                        changed: false,
-                        timedOut: false,
-                        newEvents: [],
-                    }));
-                    return;
-                }
-                // Subscribe BEFORE listFrom: any event appended after this point
-                // reaches the listener, closing the listFrom/subscribe race.
-                unsubscribe = events.subscribe(operationId, settleWithEvent);
-                const listed = events.listFrom(operationId, afterSeq);
-                if ("error" in listed) {
-                    finish(new OperationWaitError("EVENT_CURSOR_EXPIRED", "event cursor expired; re-read operation snapshot"));
-                    return;
-                }
-                if (listed.events.length > 0) {
-                    finish(settledSummary({
-                        operationId,
-                        operation: current ?? operation,
-                        afterSeq,
-                        changed: true,
-                        timedOut: false,
-                        newEvents: listed.events,
-                    }));
-                    return;
-                }
-                // No events: arm the bounded wait; the listener settles on the
-                // first new event/state change, the timer settles on timeout.
-                timer = setTimeout(() => {
-                    void (async () => {
-                        try {
-                            const latest = await input.operations.get(operationId);
-                            finish(settledSummary({
-                                operationId,
-                                operation: latest ?? operation,
-                                afterSeq,
-                                changed: false,
-                                timedOut: true,
-                                newEvents: [],
-                            }));
-                        }
-                        catch (error) {
-                            finish(error instanceof OperationWaitError
-                                ? error
-                                : new OperationWaitError("INVALID_INPUT", error instanceof Error ? error.message : String(error)));
-                        }
-                    })();
-                }, maxWaitMs);
-                timer.unref?.();
-            }
-            catch (error) {
-                finish(error instanceof OperationWaitError
-                    ? error
-                    : new OperationWaitError("INVALID_INPUT", error instanceof Error ? error.message : String(error)));
-            }
-        })();
-    });
-}