helloagents 3.0.17-beta.1 → 3.0.18-beta.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.claude-plugin/plugin.json +1 -1
- package/.codex-plugin/plugin.json +1 -1
- package/README.md +6 -17
- package/README_CN.md +6 -17
- package/bootstrap-lite.md +1 -1
- package/bootstrap.md +1 -1
- package/gemini-extension.json +1 -1
- package/package.json +1 -1
- package/scripts/cli-config.mjs +1 -0
- package/scripts/notify-events.mjs +8 -0
- package/scripts/notify-route.mjs +14 -2
- package/scripts/notify.mjs +5 -10
- package/skills/commands/init/SKILL.md +4 -3
- package/skills/commands/wiki/SKILL.md +3 -3
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "helloagents",
|
|
3
|
-
"version": "3.0.
|
|
3
|
+
"version": "3.0.18-beta.1",
|
|
4
4
|
"description": "HelloAGENTS — The orchestration kernel that makes any AI CLI smarter. Adds intelligent routing, quality verification (Ralph Loop), safety guards, and notifications.",
|
|
5
5
|
"author": {
|
|
6
6
|
"name": "HelloWind",
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "helloagents",
|
|
3
|
-
"version": "3.0.
|
|
3
|
+
"version": "3.0.18-beta.1",
|
|
4
4
|
"description": "HelloAGENTS — Quality-driven orchestration kernel for AI CLIs with intelligent routing, quality verification (Ralph Loop), safety guards, and notifications.",
|
|
5
5
|
"author": {
|
|
6
6
|
"name": "HelloWind",
|
package/README.md
CHANGED
|
@@ -8,7 +8,7 @@
|
|
|
8
8
|
|
|
9
9
|
**A workflow layer for AI coding CLIs: skills, project knowledge, delivery checks, safer config writes, and resumable execution.**
|
|
10
10
|
|
|
11
|
-
[](./package.json)
|
|
12
12
|
[](https://www.npmjs.com/package/helloagents)
|
|
13
13
|
[](./package.json)
|
|
14
14
|
[](./skills)
|
|
@@ -30,7 +30,6 @@
|
|
|
30
30
|
## Contents
|
|
31
31
|
|
|
32
32
|
- [What HelloAGENTS Does](#what-helloagents-does)
|
|
33
|
-
- [What Changed Since v3.0.16](#what-changed-since-v3016)
|
|
34
33
|
- [Core Features](#core-features)
|
|
35
34
|
- [Quick Start](#quick-start)
|
|
36
35
|
- [CLI Management](#cli-management)
|
|
@@ -77,16 +76,6 @@ HelloAGENTS adds a workflow layer on top of Claude Code, Gemini CLI, and Codex C
|
|
|
77
76
|
| Completion is vague | Natural language says “done” | Delivery checks use state, evidence, and verification |
|
|
78
77
|
| Config writes are risky | CLI files can drift | Install, update, cleanup, and doctor flows check managed files |
|
|
79
78
|
|
|
80
|
-
## What Changed Since v3.0.16
|
|
81
|
-
|
|
82
|
-
These are the main user-visible changes in `v3.0.17`, compared with `v3.0.16 beta`:
|
|
83
|
-
|
|
84
|
-
- Managed host configuration now uses `helloagents-js.cmd` as the single stable entrypoint across Windows, macOS, and Linux.
|
|
85
|
-
- Codex `notify` now writes `notify = ["helloagents-js.cmd", "codex-notify"]`, so a synced `config.toml` no longer needs OS-specific command names.
|
|
86
|
-
- Claude Code and Gemini CLI standby hooks now use the same managed entrypoint as Codex, keeping host configuration wording and behavior consistent.
|
|
87
|
-
- The package exposes `helloagents-js.cmd` as an npm bin alias while keeping the runtime command path independent of Node global package locations.
|
|
88
|
-
- Runtime-root refresh now retries transient Windows filesystem rename errors, reducing occasional update/install failures when files are still being released.
|
|
89
|
-
|
|
90
79
|
## Core Features
|
|
91
80
|
|
|
92
81
|
### 1) 14 task-aware quality skills
|
|
@@ -127,7 +116,7 @@ Commands run inside the AI CLI chat with a `~` prefix. The command skill is read
|
|
|
127
116
|
| `~prd` | Modern product requirements document through guided dimension-by-dimension exploration |
|
|
128
117
|
| `~loop` | Iterative improvement with metric, guard command, keep/revert decisions |
|
|
129
118
|
| `~wiki` | Create or sync only the project knowledge base |
|
|
130
|
-
| `~init` | Full project bootstrap: knowledge base plus project-level rule files and
|
|
119
|
+
| `~init` | Full project bootstrap: knowledge base plus project-level rule files and package-root links |
|
|
131
120
|
| `~test` | Write tests for a target module or recent change |
|
|
132
121
|
| `~verify` | Review, run verification commands, fix failures, and close out |
|
|
133
122
|
| `~commit` | Generate a conventional commit message and sync knowledge |
|
|
@@ -159,7 +148,7 @@ The knowledge base helps future turns understand the repo without re-discovering
|
|
|
159
148
|
|
|
160
149
|
`~wiki` creates or updates the knowledge base only.
|
|
161
150
|
|
|
162
|
-
`~init` does more: it creates or updates the knowledge base, writes project-level rule files, and refreshes
|
|
151
|
+
`~init` does more: it creates or updates the knowledge base, writes project-level rule files, and refreshes project-level HelloAGENTS package-root links for supported hosts.
|
|
163
152
|
|
|
164
153
|
### 4) Structured plan packages
|
|
165
154
|
|
|
@@ -523,7 +512,7 @@ This only stores short-lived `capsule.json`, `events.jsonl`, and `artifacts/`. I
|
|
|
523
512
|
| Command or setting | Behavior |
|
|
524
513
|
|--------------------|----------|
|
|
525
514
|
| `~wiki` | creates or syncs the knowledge base only |
|
|
526
|
-
| `~init` | creates knowledge base plus project-level rule files and
|
|
515
|
+
| `~init` | creates knowledge base plus project-level rule files and package-root links |
|
|
527
516
|
| `kb_create_mode = 0` | disables automatic knowledge updates |
|
|
528
517
|
| `kb_create_mode = 1` | updates knowledge automatically for coding tasks in activated projects or global mode |
|
|
529
518
|
| `kb_create_mode = 2` | updates knowledge more aggressively in activated projects or global mode |
|
|
@@ -589,7 +578,7 @@ Config file:
|
|
|
589
578
|
~/.helloagents/helloagents.json
|
|
590
579
|
```
|
|
591
580
|
|
|
592
|
-
Default shape
|
|
581
|
+
Default shape:
|
|
593
582
|
|
|
594
583
|
```json
|
|
595
584
|
{
|
|
@@ -684,7 +673,7 @@ Both.
|
|
|
684
673
|
|
|
685
674
|
Use `~wiki` when you only want project knowledge.
|
|
686
675
|
|
|
687
|
-
Use `~init` when you also want project-level rule files and
|
|
676
|
+
Use `~init` when you also want project-level rule files and project-level HelloAGENTS package-root links.
|
|
688
677
|
|
|
689
678
|
### What is the difference between standby and global?
|
|
690
679
|
|
package/README_CN.md
CHANGED
|
@@ -8,7 +8,7 @@
|
|
|
8
8
|
|
|
9
9
|
**面向 AI 编码 CLI 的工作流层:技能、知识库、交付检查、更安全的配置写入,以及可恢复的执行流程。**
|
|
10
10
|
|
|
11
|
-
[](./package.json)
|
|
12
12
|
[](https://www.npmjs.com/package/helloagents)
|
|
13
13
|
[](./package.json)
|
|
14
14
|
[](./skills)
|
|
@@ -30,7 +30,6 @@
|
|
|
30
30
|
## 目录
|
|
31
31
|
|
|
32
32
|
- [HelloAGENTS 做什么](#helloagents-做什么)
|
|
33
|
-
- [相对 v3.0.16 的真实变化](#相对-v3016-的真实变化)
|
|
34
33
|
- [核心功能](#核心功能)
|
|
35
34
|
- [快速开始](#快速开始)
|
|
36
35
|
- [CLI 管理](#cli-管理)
|
|
@@ -77,16 +76,6 @@ HelloAGENTS 叠加在 Claude Code、Gemini CLI 和 Codex CLI 之上,帮助模
|
|
|
77
76
|
| 完成态模糊 | 自然语言说“完成” | 按状态、证据和验证结果交付 |
|
|
78
77
|
| 配置容易漂移 | CLI 文件可能不一致 | 安装、更新、清理和 doctor 会检查受管文件 |
|
|
79
78
|
|
|
80
|
-
## 相对 v3.0.16 的真实变化
|
|
81
|
-
|
|
82
|
-
下面是当前 `v3.0.17` 相对 `v3.0.16 beta` 的主要可见变化:
|
|
83
|
-
|
|
84
|
-
- 受管宿主配置统一使用 `helloagents-js.cmd` 作为 Windows、macOS 和 Linux 三端稳定入口。
|
|
85
|
-
- Codex `notify` 统一写入 `notify = ["helloagents-js.cmd", "codex-notify"]`,同步同一份 `config.toml` 时不再需要区分系统。
|
|
86
|
-
- Claude Code 与 Gemini CLI 的标准模式 hooks 也使用同一受管入口,与 Codex 的配置行为和文档表述保持一致。
|
|
87
|
-
- npm 包新增 `helloagents-js.cmd` bin alias,同时保持运行命令不依赖 Node 全局包的实际安装路径。
|
|
88
|
-
- 稳定运行根目录刷新增加 Windows 临时文件系统重试,减少文件释放延迟导致的偶发更新/安装失败。
|
|
89
|
-
|
|
90
79
|
## 核心功能
|
|
91
80
|
|
|
92
81
|
### 1)14 个按任务使用的质量技能
|
|
@@ -127,7 +116,7 @@ HelloAGENTS 内置 14 个 `hello-*` 技能。技能只在当前阶段需要时
|
|
|
127
116
|
| `~prd` | 通过逐维度讨论生成现代产品需求文档 |
|
|
128
117
|
| `~loop` | 设置指标和守卫命令,循环改进、保留或回滚 |
|
|
129
118
|
| `~wiki` | 只创建或同步项目知识库 |
|
|
130
|
-
| `~init` |
|
|
119
|
+
| `~init` | 完整项目初始化:知识库、项目级规则文件和包根链接 |
|
|
131
120
|
| `~test` | 为指定模块或最近变更编写测试 |
|
|
132
121
|
| `~verify` | 审查、运行验证命令、修复失败并收尾 |
|
|
133
122
|
| `~commit` | 生成规范化提交信息并同步知识库 |
|
|
@@ -159,7 +148,7 @@ HelloAGENTS 可以在 `.helloagents/` 下创建和维护项目知识库。
|
|
|
159
148
|
|
|
160
149
|
`~wiki` 只创建或更新知识库。
|
|
161
150
|
|
|
162
|
-
`~init`
|
|
151
|
+
`~init` 做得更多:创建或更新知识库、写入项目级规则文件,并刷新各宿主项目级 HelloAGENTS 包根链接。
|
|
163
152
|
|
|
164
153
|
### 4)结构化方案包
|
|
165
154
|
|
|
@@ -525,7 +514,7 @@ Codex 全局模式由 HelloAGENTS 通过本地插件路径自动安装。
|
|
|
525
514
|
| 命令或配置 | 行为 |
|
|
526
515
|
|------------|------|
|
|
527
516
|
| `~wiki` | 只创建或同步知识库 |
|
|
528
|
-
| `~init` |
|
|
517
|
+
| `~init` | 创建知识库,同时写入项目级规则文件和包根链接 |
|
|
529
518
|
| `kb_create_mode = 0` | 关闭自动知识更新 |
|
|
530
519
|
| `kb_create_mode = 1` | 已激活项目或全局模式中,编码任务自动更新知识 |
|
|
531
520
|
| `kb_create_mode = 2` | 已激活项目或全局模式中,更积极地更新知识 |
|
|
@@ -591,7 +580,7 @@ UI 任务遵循以下优先级:
|
|
|
591
580
|
~/.helloagents/helloagents.json
|
|
592
581
|
```
|
|
593
582
|
|
|
594
|
-
|
|
583
|
+
默认结构:
|
|
595
584
|
|
|
596
585
|
```json
|
|
597
586
|
{
|
|
@@ -686,7 +675,7 @@ npm test
|
|
|
686
675
|
|
|
687
676
|
只想创建项目知识库,用 `~wiki`。
|
|
688
677
|
|
|
689
|
-
|
|
678
|
+
还想写入项目级规则文件和宿主项目级 HelloAGENTS 包根链接,用 `~init`。
|
|
690
679
|
|
|
691
680
|
### standby 和 global 有什么区别?
|
|
692
681
|
|
package/bootstrap-lite.md
CHANGED
|
@@ -116,7 +116,7 @@
|
|
|
116
116
|
图标:💡直接响应 | ⚡快速执行 | 🔵规划流程 | ✅完成 | ❓等待输入 | ⚠️警告 | ❌错误
|
|
117
117
|
|
|
118
118
|
使用约束:
|
|
119
|
-
-
|
|
119
|
+
- 首行必须保留 `【HelloAGENTS】` 和连字符 `-`,不得省略;状态图标与收尾内容必须一致。正文仍在等待用户输入、确认、授权或补充信息(含确认是否执行已给出的方案)时,只能使用 `❓等待输入`;仅在本轮执行已完成且不存在待确认动作时,才能使用 `✅完成`。
|
|
120
120
|
- `🔄 下一步` 必须写真实下一步,不改写成条件式能力表述或空泛询问。若正在等待确认,写清待确认动作;若存在自然后续动作,直接给出明确引导;若当前任务已完整结束且确无合理后续,可填写“当前任务已完成;无后续动作。”
|
|
121
121
|
|
|
122
122
|
### 收尾状态信号
|
package/bootstrap.md
CHANGED
|
@@ -116,7 +116,7 @@
|
|
|
116
116
|
图标:💡直接响应 | ⚡快速执行 | 🔵规划流程 | ✅完成 | ❓等待输入 | ⚠️警告 | ❌错误
|
|
117
117
|
|
|
118
118
|
使用约束:
|
|
119
|
-
-
|
|
119
|
+
- 首行必须保留 `【HelloAGENTS】` 和连字符 `-`,不得省略;状态图标与收尾内容必须一致。正文仍在等待用户输入、确认、授权或补充信息(含确认是否执行已给出的方案)时,只能使用 `❓等待输入`;仅在本轮执行已完成且不存在待确认动作时,才能使用 `✅完成`。
|
|
120
120
|
- `🔄 下一步` 必须写真实下一步,不改写成条件式能力表述或空泛询问。若正在等待确认,写清待确认动作;若存在自然后续动作,直接给出明确引导;若当前任务已完整结束且确无合理后续,可填写“当前任务已完成;无后续动作。”
|
|
121
121
|
|
|
122
122
|
### 收尾状态信号
|
package/gemini-extension.json
CHANGED
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "helloagents",
|
|
3
|
-
"version": "3.0.
|
|
3
|
+
"version": "3.0.18-beta.1",
|
|
4
4
|
"type": "module",
|
|
5
5
|
"description": "HelloAGENTS — The orchestration kernel that makes any AI CLI smarter. Adds intelligent routing, quality verification (Ralph Loop), safety guards, and notifications.",
|
|
6
6
|
"author": "HelloWind",
|
package/scripts/cli-config.mjs
CHANGED
|
@@ -7,3 +7,11 @@ export function shouldIgnoreCodexNotifyClient(client) {
|
|
|
7
7
|
export function shouldIgnoreFormattedSubagent(lastMsg, outputFormatEnabled) {
|
|
8
8
|
return outputFormatEnabled && !lastMsg.includes('【HelloAGENTS】');
|
|
9
9
|
}
|
|
10
|
+
|
|
11
|
+
export function resolveNotifyHost(argv = []) {
|
|
12
|
+
const args = Array.from(argv, (value) => String(value || ''));
|
|
13
|
+
const command = args[2] || args[0] || '';
|
|
14
|
+
if (args.includes('--gemini')) return 'gemini';
|
|
15
|
+
if (args.includes('--codex') || command === 'codex-notify') return 'codex';
|
|
16
|
+
return 'claude';
|
|
17
|
+
}
|
package/scripts/notify-route.mjs
CHANGED
|
@@ -1,6 +1,18 @@
|
|
|
1
1
|
import { isProjectRuntimeActive } from './runtime-scope.mjs'
|
|
2
2
|
|
|
3
|
-
export function
|
|
3
|
+
export function resolveRuntimeInstallMode(settings = {}, host = '') {
|
|
4
|
+
if (!settings || typeof settings !== 'object') return 'standby'
|
|
5
|
+
const hostModes = settings.host_install_modes
|
|
6
|
+
const trackedMode = hostModes && typeof hostModes === 'object' && !Array.isArray(hostModes)
|
|
7
|
+
? hostModes[host] || ''
|
|
8
|
+
: ''
|
|
9
|
+
return trackedMode || settings.install_mode || 'standby'
|
|
10
|
+
}
|
|
11
|
+
|
|
12
|
+
export function resolveBootstrapFile(cwd, settings = {}, host = '') {
|
|
13
|
+
const installMode = typeof settings === 'string'
|
|
14
|
+
? settings
|
|
15
|
+
: resolveRuntimeInstallMode(settings, host)
|
|
4
16
|
const isActivated = isProjectRuntimeActive(cwd)
|
|
5
17
|
return (installMode === 'global' || isActivated) ? 'bootstrap.md' : 'bootstrap-lite.md'
|
|
6
18
|
}
|
|
@@ -97,7 +109,7 @@ export function handleRouteCommand({
|
|
|
97
109
|
return
|
|
98
110
|
}
|
|
99
111
|
|
|
100
|
-
const bootstrapFile = resolveBootstrapFile(cwd, settings
|
|
112
|
+
const bootstrapFile = resolveBootstrapFile(cwd, settings, host)
|
|
101
113
|
if (bootstrapFile === 'bootstrap.md') {
|
|
102
114
|
clearRouteContext({ cwd, payload })
|
|
103
115
|
appendReplayEvent(cwd, {
|
package/scripts/notify.mjs
CHANGED
|
@@ -9,7 +9,7 @@ import { homedir } from 'node:os';
|
|
|
9
9
|
import { playSound as _playSound, desktopNotify as _desktopNotify } from './notify-ui.mjs';
|
|
10
10
|
import { resolveNotificationSource } from './notify-source.mjs';
|
|
11
11
|
import { buildCompactionContext, buildInjectContext, buildRouteInstruction, buildSemanticRouteInstruction, resolveCanonicalCommandSkill } from './notify-context.mjs';
|
|
12
|
-
import { shouldIgnoreCodexNotifyClient } from './notify-events.mjs';
|
|
12
|
+
import { resolveNotifyHost, shouldIgnoreCodexNotifyClient } from './notify-events.mjs';
|
|
13
13
|
import { runGateScript } from './notify-gates.mjs';
|
|
14
14
|
import { handleRouteCommand, resolveBootstrapFile } from './notify-route.mjs';
|
|
15
15
|
import { readSettings, readStdinJson, output, suppressedOutput, emptySuppress } from './notify-shared.mjs';
|
|
@@ -24,11 +24,8 @@ const __filename = fileURLToPath(import.meta.url);
|
|
|
24
24
|
const __dirname = dirname(__filename);
|
|
25
25
|
const PKG_ROOT = join(__dirname, '..');
|
|
26
26
|
const CONFIG_FILE = join(homedir(), '.helloagents', 'helloagents.json');
|
|
27
|
-
const
|
|
28
|
-
|
|
29
|
-
: process.argv.includes('--codex')
|
|
30
|
-
? 'codex'
|
|
31
|
-
: 'claude';
|
|
27
|
+
const cmd = process.argv[2] || '';
|
|
28
|
+
const HOST = resolveNotifyHost(process.argv);
|
|
32
29
|
const IS_GEMINI = HOST === 'gemini';
|
|
33
30
|
const EVENT_NAME = {
|
|
34
31
|
SessionStart: 'SessionStart',
|
|
@@ -151,7 +148,7 @@ function cmdPreCompact() {
|
|
|
151
148
|
const payload = readStdinJson();
|
|
152
149
|
const cwd = payload.cwd || process.cwd();
|
|
153
150
|
const settings = getSettings();
|
|
154
|
-
const bootstrapFile = resolveBootstrapFile(cwd, settings
|
|
151
|
+
const bootstrapFile = resolveBootstrapFile(cwd, settings, HOST);
|
|
155
152
|
const context = buildCompactionContext({
|
|
156
153
|
payload,
|
|
157
154
|
pkgRoot: PKG_ROOT,
|
|
@@ -197,7 +194,7 @@ function cmdInject() {
|
|
|
197
194
|
const source = payload.source || 'startup';
|
|
198
195
|
const cwd = payload.cwd || process.cwd();
|
|
199
196
|
const settings = getSettings();
|
|
200
|
-
const bootstrapFile = resolveBootstrapFile(cwd, settings
|
|
197
|
+
const bootstrapFile = resolveBootstrapFile(cwd, settings, HOST);
|
|
201
198
|
|
|
202
199
|
let bootstrap = '';
|
|
203
200
|
try {
|
|
@@ -323,8 +320,6 @@ function cmdCodexNotify() {
|
|
|
323
320
|
clearRouteContext({ cwd, payload: turnPayload });
|
|
324
321
|
}
|
|
325
322
|
|
|
326
|
-
const cmd = process.argv[2] || '';
|
|
327
|
-
|
|
328
323
|
switch (cmd) {
|
|
329
324
|
case 'inject': cmdInject(); break;
|
|
330
325
|
case 'stop': cmdStop(); break;
|
|
@@ -7,7 +7,7 @@ policy:
|
|
|
7
7
|
Trigger: ~init
|
|
8
8
|
|
|
9
9
|
~init 是用户显式命令,创建完整知识库,不受 kb_create_mode 限制。
|
|
10
|
-
执行 `~init` 时,`.helloagents/` 目录结构、模板格式和状态文件规则按当前已加载 bootstrap
|
|
10
|
+
执行 `~init` 时,`.helloagents/` 目录结构、模板格式和状态文件规则按当前已加载 bootstrap 执行;本命令额外负责项目级规则文件和各宿主项目级 HelloAGENTS 包根链接。
|
|
11
11
|
`.helloagents/` 在本 skill 中统一按项目级存储路径理解:项目本地 `.helloagents/` 继续承担激活目录;状态文件只使用 `state_path`;若 `project_store_mode=repo-shared`,知识库、`DESIGN.md` 与方案包按当前上下文中已注入的项目知识/方案目录写入。
|
|
12
12
|
|
|
13
13
|
## 流程
|
|
@@ -16,10 +16,11 @@ Trigger: ~init
|
|
|
16
16
|
|
|
17
17
|
1. 创建 `.helloagents/` 目录 + `state_path`(按 templates/STATE.md 格式,初始“主线目标”写当前初始化任务,初始状态为空闲)
|
|
18
18
|
2. 定位插件根目录:优先读取当前上下文中已注入的“当前 HelloAGENTS 包根目录”;若上下文未提供,再根据当前已加载的规则文件反推,禁止猜测其他目录
|
|
19
|
-
3.
|
|
19
|
+
3. 刷新各宿主项目级 HelloAGENTS 包根链接(删除旧的重建):
|
|
20
20
|
- `.claude/skills/helloagents` symlink → `{插件根目录}/`
|
|
21
21
|
- `.gemini/skills/helloagents` symlink → `{插件根目录}/`
|
|
22
22
|
- `.codex/skills/helloagents` symlink → `{插件根目录}/`
|
|
23
|
+
这些链接用于项目级规则定位 HelloAGENTS 的 `skills/`、`templates/` 和 `scripts/`;宿主若支持递归发现 `SKILL.md`,也可直接识别包内 skills。
|
|
23
24
|
4. 读取 `{插件根目录}/bootstrap.md`,用 `<!-- HELLOAGENTS_START -->` / `<!-- HELLOAGENTS_END -->` 标记包裹后写入:
|
|
24
25
|
- `AGENTS.md`(项目根目录,Codex 读取)
|
|
25
26
|
- `CLAUDE.md`(项目根目录,Claude Code 读取)
|
|
@@ -63,6 +64,6 @@ commands:
|
|
|
63
64
|
重复执行 ~init 是安全的:
|
|
64
65
|
- 已存在的 .helloagents/ 文件不覆盖
|
|
65
66
|
- `state_path` 只记录当前初始化任务;后续进入其他任务时必须按新任务重写
|
|
66
|
-
-
|
|
67
|
+
- 各宿主项目级 HelloAGENTS 包根链接会刷新(删除旧的重建)
|
|
67
68
|
- AGENTS.md/CLAUDE.md/GEMINI.md 中标记内容替换更新
|
|
68
69
|
- .gitignore 只追加缺失行
|
|
@@ -9,7 +9,7 @@ Trigger: ~wiki
|
|
|
9
9
|
`~wiki` 是用户显式命令,仅创建、补全或同步项目知识库。
|
|
10
10
|
|
|
11
11
|
`~wiki` 是显式知识库命令,不受 `kb_create_mode` 限制。
|
|
12
|
-
执行 `~wiki` 时,`.helloagents/` 目录结构、模板格式和状态文件重写规则按当前已加载 bootstrap
|
|
12
|
+
执行 `~wiki` 时,`.helloagents/` 目录结构、模板格式和状态文件重写规则按当前已加载 bootstrap 执行;不写入项目级规则文件,也不创建项目级 HelloAGENTS 包根链接。
|
|
13
13
|
`.helloagents/` 在本 skill 中统一按项目级存储路径理解:状态文件只使用 `state_path`;若 `project_store_mode=repo-shared`,`context.md`、`guidelines.md`、`verify.yaml`、`CHANGELOG.md`、`DESIGN.md`、`modules/` 改按当前上下文中已注入的项目知识目录写入。
|
|
14
14
|
|
|
15
15
|
## 流程
|
|
@@ -23,7 +23,7 @@ Trigger: ~wiki
|
|
|
23
23
|
```
|
|
24
24
|
3. 明确不执行以下操作:
|
|
25
25
|
- 不创建或更新项目级规则文件(`AGENTS.md`、`CLAUDE.md`、`.gemini/GEMINI.md`)
|
|
26
|
-
-
|
|
26
|
+
- 不创建项目级 HelloAGENTS 包根链接
|
|
27
27
|
|
|
28
28
|
### 阶段 2:知识库创建或补全(条件性)
|
|
29
29
|
|
|
@@ -54,4 +54,4 @@ commands:
|
|
|
54
54
|
- `state_path` 按当前任务状态重写,不追加历史;它只记录当前知识库任务,不承担项目的长期记忆
|
|
55
55
|
- 知识库文件缺失时补全,已存在时按模板增量更新
|
|
56
56
|
- `.gitignore` 只追加缺失行
|
|
57
|
-
-
|
|
57
|
+
- 永不写入项目级规则文件,也不创建任何项目级 HelloAGENTS 包根链接
|