@evandu/super-pi-server 0.0.0-stage → 0.1.259
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/README.md +248 -3
- package/agents/code-auditor.md +64 -0
- package/agents/code-designer.md +64 -0
- package/agents/coder.md +24 -0
- package/agents/critic.md +24 -0
- package/agents/document-organizer.md +48 -0
- package/agents/patcher.md +27 -0
- package/agents/planner.md +30 -0
- package/agents/project-leader.md +28 -0
- package/agents/reviewer.md +27 -0
- package/agents/scout.md +27 -0
- package/agents/solution-designer.md +62 -0
- package/agents/super-pi-documents/index.ts +92 -0
- package/agents/super-pi-feishu/index.ts +88 -0
- package/agents/super-pi-knowledge/index.ts +70 -0
- package/agents/super-pi-managed-model/index.ts +9 -0
- package/agents/super-pi-memory/index.ts +75 -0
- package/agents/super-pi-native-tools/index.ts +1 -0
- package/agents/super-pi-task/core.ts +2262 -0
- package/agents/super-pi-task/index.ts +814 -0
- package/agents/super-pi-task/workflow-schema.ts +34 -0
- package/agents/super-pi-web/index.ts +32 -0
- package/agents/verifier.md +25 -0
- package/assets/fonts/LICENSE +92 -0
- package/assets/fonts/NotoSansSC-Bold.otf +0 -0
- package/assets/fonts/NotoSansSC-Regular.otf +0 -0
- package/assets/knowledge-model/README.md +37 -0
- package/assets/knowledge-model/Xenova/bge-small-zh-v1.5/config.json +31 -0
- package/assets/knowledge-model/Xenova/bge-small-zh-v1.5/onnx/model_quantized.onnx +0 -0
- package/assets/knowledge-model/Xenova/bge-small-zh-v1.5/special_tokens_map.json +7 -0
- package/assets/knowledge-model/Xenova/bge-small-zh-v1.5/tokenizer.json +21278 -0
- package/assets/knowledge-model/Xenova/bge-small-zh-v1.5/tokenizer_config.json +15 -0
- package/assets/knowledge-model/Xenova/bge-small-zh-v1.5/vocab.txt +21128 -0
- package/bin/npm-prefix.d.ts +1 -0
- package/bin/npm-prefix.js +10 -0
- package/bin/super-pi.js +349 -0
- package/dist/client/.super-pi-client-sources.json +82 -0
- package/dist/client/assets/CodeMirrorEditor-mi9rGUce.js +2 -0
- package/dist/client/assets/GitDiffEditor-CpoDDQrA.css +1 -0
- package/dist/client/assets/GitDiffEditor-FQBfGxOY.js +29 -0
- package/dist/client/assets/diff-ChtP43wD.js +1 -0
- package/dist/client/assets/dist-AW5lyI3I.js +1 -0
- package/dist/client/assets/dist-B3-a6uF0.js +7 -0
- package/dist/client/assets/dist-BGxkyh6e.js +1 -0
- package/dist/client/assets/dist-BJxfd7yb.js +11 -0
- package/dist/client/assets/dist-BRVQKlcp.js +1 -0
- package/dist/client/assets/dist-BV64A9Ji.js +1 -0
- package/dist/client/assets/dist-BXAOrsjq.js +1 -0
- package/dist/client/assets/dist-CKF2Jksy.js +1 -0
- package/dist/client/assets/dist-CX_Hm4GS.js +1 -0
- package/dist/client/assets/dist-Cus6_oaF.js +2 -0
- package/dist/client/assets/dist-Cx00EZKR.js +23 -0
- package/dist/client/assets/dist-D71y2EI1.js +1 -0
- package/dist/client/assets/dist-D978o95T.js +5 -0
- package/dist/client/assets/dist-Dc1MBGAG.js +1 -0
- package/dist/client/assets/dist-Dds-qLxY.js +9 -0
- package/dist/client/assets/dist-DeD9copn.js +1 -0
- package/dist/client/assets/dist-DuvGReu-.js +6 -0
- package/dist/client/assets/dist-Dvb0SqcK.js +2 -0
- package/dist/client/assets/dist-J6XWqRpB.js +3 -0
- package/dist/client/assets/dist-QzbOCAU6.js +1 -0
- package/dist/client/assets/dist-moiMRpw_.js +1 -0
- package/dist/client/assets/dist-scBkrwv-.js +2 -0
- package/dist/client/assets/dockerfile-BDePf5PZ.js +1 -0
- package/dist/client/assets/index-Cn7hdrDw.js +350 -0
- package/dist/client/assets/index-UfzqdECL.css +1 -0
- package/dist/client/assets/powershell-D8WpQPaX.js +1 -0
- package/dist/client/assets/properties-C357ku5U.js +1 -0
- package/dist/client/assets/rolldown-runtime-XKyaHtl9.js +1 -0
- package/dist/client/assets/ruby-ChfNwlCY.js +1 -0
- package/dist/client/assets/shell-DwuoZtxw.js +1 -0
- package/dist/client/assets/swift-DKB6_1j6.js +1 -0
- package/dist/client/assets/toml-BPTmHmyx.js +1 -0
- package/dist/client/favicon.ico +0 -0
- package/dist/client/favicon.png +0 -0
- package/dist/client/favicon.svg +10 -0
- package/dist/client/index.html +18 -0
- package/dist/extension.d.ts +15 -0
- package/dist/extension.js +126 -0
- package/dist/server/active-memory.d.ts +247 -0
- package/dist/server/active-memory.js +669 -0
- package/dist/server/agents.d.ts +245 -0
- package/dist/server/agents.js +701 -0
- package/dist/server/application/catalog-service.d.ts +98 -0
- package/dist/server/application/catalog-service.js +245 -0
- package/dist/server/application/conversation-service.d.ts +245 -0
- package/dist/server/application/conversation-service.js +654 -0
- package/dist/server/application/document-service.d.ts +68 -0
- package/dist/server/application/document-service.js +120 -0
- package/dist/server/application/extension-ui-service.d.ts +44 -0
- package/dist/server/application/extension-ui-service.js +135 -0
- package/dist/server/application/feishu-service.d.ts +179 -0
- package/dist/server/application/feishu-service.js +358 -0
- package/dist/server/application/group-chat-service.d.ts +75 -0
- package/dist/server/application/group-chat-service.js +953 -0
- package/dist/server/application/group-execution.d.ts +15 -0
- package/dist/server/application/group-execution.js +77 -0
- package/dist/server/application/project-service.d.ts +464 -0
- package/dist/server/application/project-service.js +2523 -0
- package/dist/server/application/provider-service.d.ts +174 -0
- package/dist/server/application/provider-service.js +510 -0
- package/dist/server/application/run-service.d.ts +547 -0
- package/dist/server/application/run-service.js +2683 -0
- package/dist/server/application/scheduled-task-service.d.ts +184 -0
- package/dist/server/application/scheduled-task-service.js +391 -0
- package/dist/server/application/session-orchestrator-service.d.ts +388 -0
- package/dist/server/application/session-orchestrator-service.js +1146 -0
- package/dist/server/application/session-reminder-service.d.ts +59 -0
- package/dist/server/application/session-reminder-service.js +141 -0
- package/dist/server/application/session-watchdog.d.ts +171 -0
- package/dist/server/application/session-watchdog.js +284 -0
- package/dist/server/application/workspace-service.d.ts +52 -0
- package/dist/server/application/workspace-service.js +186 -0
- package/dist/server/application/worktree-service.d.ts +13 -0
- package/dist/server/application/worktree-service.js +76 -0
- package/dist/server/browser-access.d.ts +34 -0
- package/dist/server/browser-access.js +89 -0
- package/dist/server/capability.d.ts +105 -0
- package/dist/server/capability.js +170 -0
- package/dist/server/command-handler.d.ts +93 -0
- package/dist/server/command-handler.js +540 -0
- package/dist/server/content-search.d.ts +16 -0
- package/dist/server/content-search.js +235 -0
- package/dist/server/conversation-usage.d.ts +22 -0
- package/dist/server/conversation-usage.js +43 -0
- package/dist/server/documents.d.ts +152 -0
- package/dist/server/documents.js +998 -0
- package/dist/server/engineering-learning.d.ts +30 -0
- package/dist/server/engineering-learning.js +106 -0
- package/dist/server/exec-service.d.ts +45 -0
- package/dist/server/exec-service.js +167 -0
- package/dist/server/explicit-file-open.d.ts +34 -0
- package/dist/server/explicit-file-open.js +105 -0
- package/dist/server/failure-text.d.ts +18 -0
- package/dist/server/failure-text.js +44 -0
- package/dist/server/feishu-cli.d.ts +223 -0
- package/dist/server/feishu-cli.js +464 -0
- package/dist/server/file-changes.d.ts +119 -0
- package/dist/server/file-changes.js +280 -0
- package/dist/server/file-diff.d.ts +41 -0
- package/dist/server/file-diff.js +97 -0
- package/dist/server/file-index.d.ts +100 -0
- package/dist/server/file-index.js +296 -0
- package/dist/server/file-name-index.d.ts +160 -0
- package/dist/server/file-name-index.js +411 -0
- package/dist/server/files.d.ts +269 -0
- package/dist/server/files.js +748 -0
- package/dist/server/git.d.ts +245 -0
- package/dist/server/git.js +903 -0
- package/dist/server/http-server.d.ts +45 -0
- package/dist/server/http-server.js +3830 -0
- package/dist/server/index.d.ts +1 -0
- package/dist/server/index.js +485 -0
- package/dist/server/inline-thinking.d.ts +90 -0
- package/dist/server/inline-thinking.js +181 -0
- package/dist/server/json-comments.d.ts +14 -0
- package/dist/server/json-comments.js +71 -0
- package/dist/server/knowledge-embedding-worker.d.ts +1 -0
- package/dist/server/knowledge-embedding-worker.js +27 -0
- package/dist/server/knowledge-vectors.d.ts +150 -0
- package/dist/server/knowledge-vectors.js +385 -0
- package/dist/server/knowledge.d.ts +273 -0
- package/dist/server/knowledge.js +848 -0
- package/dist/server/length-continuation.d.ts +83 -0
- package/dist/server/length-continuation.js +118 -0
- package/dist/server/logging.d.ts +32 -0
- package/dist/server/logging.js +185 -0
- package/dist/server/managed-http-tunnel.d.ts +6 -0
- package/dist/server/managed-http-tunnel.js +85 -0
- package/dist/server/managed-http.d.ts +4 -0
- package/dist/server/managed-http.js +85 -0
- package/dist/server/managed-model-adapter.d.ts +17 -0
- package/dist/server/managed-model-adapter.js +75 -0
- package/dist/server/managed-model-gateway.d.ts +49 -0
- package/dist/server/managed-model-gateway.js +73 -0
- package/dist/server/managed-servers.d.ts +55 -0
- package/dist/server/managed-servers.js +392 -0
- package/dist/server/managed-ssh.d.ts +26 -0
- package/dist/server/managed-ssh.js +73 -0
- package/dist/server/manager-vault.d.ts +11 -0
- package/dist/server/manager-vault.js +67 -0
- package/dist/server/memory-recall.d.ts +25 -0
- package/dist/server/memory-recall.js +45 -0
- package/dist/server/memory.d.ts +67 -0
- package/dist/server/memory.js +216 -0
- package/dist/server/message-media.d.ts +7 -0
- package/dist/server/message-media.js +54 -0
- package/dist/server/model-context-window.d.ts +186 -0
- package/dist/server/model-context-window.js +444 -0
- package/dist/server/model-filter.d.ts +23 -0
- package/dist/server/model-filter.js +48 -0
- package/dist/server/models-store.d.ts +36 -0
- package/dist/server/models-store.js +132 -0
- package/dist/server/native-tool-settings.d.ts +25 -0
- package/dist/server/native-tool-settings.js +41 -0
- package/dist/server/native-tools-extension.d.ts +2 -0
- package/dist/server/native-tools-extension.js +39 -0
- package/dist/server/native-tools-service.d.ts +39 -0
- package/dist/server/native-tools-service.js +240 -0
- package/dist/server/persistence/app-state-repository.d.ts +65 -0
- package/dist/server/persistence/app-state-repository.js +1 -0
- package/dist/server/persistence/json-state-repository.d.ts +14 -0
- package/dist/server/persistence/json-state-repository.js +43 -0
- package/dist/server/persistence/open-state-store.d.ts +3 -0
- package/dist/server/persistence/open-state-store.js +35 -0
- package/dist/server/persistence/rename.d.ts +22 -0
- package/dist/server/persistence/rename.js +44 -0
- package/dist/server/persistence/sqlite-state-repository.d.ts +12 -0
- package/dist/server/persistence/sqlite-state-repository.js +264 -0
- package/dist/server/pi-command.d.ts +94 -0
- package/dist/server/pi-command.js +265 -0
- package/dist/server/pi-native.d.ts +12 -0
- package/dist/server/pi-native.js +16 -0
- package/dist/server/pi-session-events.d.ts +158 -0
- package/dist/server/pi-session-events.js +373 -0
- package/dist/server/pi-session-pool.d.ts +233 -0
- package/dist/server/pi-session-pool.js +382 -0
- package/dist/server/pi-session.d.ts +642 -0
- package/dist/server/pi-session.js +655 -0
- package/dist/server/presentations/schema.d.ts +120 -0
- package/dist/server/presentations/schema.js +128 -0
- package/dist/server/provider-network.d.ts +16 -0
- package/dist/server/provider-network.js +106 -0
- package/dist/server/provider-proxy.d.ts +18 -0
- package/dist/server/provider-proxy.js +200 -0
- package/dist/server/providers.d.ts +333 -0
- package/dist/server/providers.js +989 -0
- package/dist/server/remote-bootstrap.d.ts +17 -0
- package/dist/server/remote-bootstrap.js +119 -0
- package/dist/server/rpc-channel.d.ts +131 -0
- package/dist/server/rpc-channel.js +135 -0
- package/dist/server/runtime.d.ts +1393 -0
- package/dist/server/runtime.js +3286 -0
- package/dist/server/scheduled-tasks.d.ts +101 -0
- package/dist/server/scheduled-tasks.js +138 -0
- package/dist/server/server-context.d.ts +119 -0
- package/dist/server/server-context.js +1 -0
- package/dist/server/session-cwd.d.ts +19 -0
- package/dist/server/session-cwd.js +58 -0
- package/dist/server/session-ledger.d.ts +67 -0
- package/dist/server/session-ledger.js +76 -0
- package/dist/server/session-memory-scope.d.ts +4 -0
- package/dist/server/session-memory-scope.js +30 -0
- package/dist/server/session-repository.d.ts +100 -0
- package/dist/server/session-repository.js +246 -0
- package/dist/server/session-tree-scope.d.ts +125 -0
- package/dist/server/session-tree-scope.js +337 -0
- package/dist/server/skills.d.ts +23 -0
- package/dist/server/skills.js +190 -0
- package/dist/server/store.d.ts +1062 -0
- package/dist/server/store.js +3305 -0
- package/dist/server/subagent-runs.d.ts +185 -0
- package/dist/server/subagent-runs.js +383 -0
- package/dist/server/subagents.d.ts +200 -0
- package/dist/server/subagents.js +619 -0
- package/dist/server/task-tool.d.ts +282 -0
- package/dist/server/task-tool.js +507 -0
- package/dist/server/tavily.d.ts +32 -0
- package/dist/server/tavily.js +136 -0
- package/dist/server/team-context.d.ts +12 -0
- package/dist/server/team-context.js +180 -0
- package/dist/server/terminal.d.ts +163 -0
- package/dist/server/terminal.js +557 -0
- package/dist/server/thinking-level-discovery.d.ts +38 -0
- package/dist/server/thinking-level-discovery.js +183 -0
- package/dist/server/time-zone.d.ts +20 -0
- package/dist/server/time-zone.js +32 -0
- package/dist/server/tool-calls.d.ts +130 -0
- package/dist/server/tool-calls.js +341 -0
- package/dist/server/tool-llm.d.ts +72 -0
- package/dist/server/tool-llm.js +146 -0
- package/dist/server/tool-registry.d.ts +129 -0
- package/dist/server/tool-registry.js +249 -0
- package/dist/server/tool-runner.d.ts +112 -0
- package/dist/server/tool-runner.js +306 -0
- package/dist/server/turn-failure.d.ts +48 -0
- package/dist/server/turn-failure.js +148 -0
- package/dist/server/types.d.ts +2315 -0
- package/dist/server/types.js +11 -0
- package/dist/server/update-status.d.ts +14 -0
- package/dist/server/update-status.js +69 -0
- package/dist/server/update.d.ts +46 -0
- package/dist/server/update.js +104 -0
- package/dist/server/upload.d.ts +35 -0
- package/dist/server/upload.js +177 -0
- package/dist/server/version.d.ts +33 -0
- package/dist/server/version.js +111 -0
- package/dist/server/web.d.ts +27 -0
- package/dist/server/web.js +183 -0
- package/dist/server/workflow-engine.d.ts +47 -0
- package/dist/server/workflow-engine.js +255 -0
- package/dist/server/workflow.d.ts +329 -0
- package/dist/server/workflow.js +469 -0
- package/dist/server/workspace-file-indexes.d.ts +2 -0
- package/dist/server/workspace-file-indexes.js +61 -0
- package/dist/server/workspace-file-operations.d.ts +36 -0
- package/dist/server/workspace-file-operations.js +281 -0
- package/dist/server/workspace-file-response.d.ts +7 -0
- package/dist/server/workspace-file-response.js +79 -0
- package/dist/server/ws-server.d.ts +149 -0
- package/dist/server/ws-server.js +640 -0
- package/dist/shared/agent-order.d.ts +44 -0
- package/dist/shared/agent-order.js +50 -0
- package/dist/shared/archive-decisions.d.ts +44 -0
- package/dist/shared/archive-decisions.js +98 -0
- package/dist/shared/background-task.d.ts +25 -0
- package/dist/shared/background-task.js +183 -0
- package/dist/shared/child-session-lock.d.ts +37 -0
- package/dist/shared/child-session-lock.js +59 -0
- package/dist/shared/content-search.d.ts +22 -0
- package/dist/shared/content-search.js +1 -0
- package/dist/shared/conversation-title.d.ts +4 -0
- package/dist/shared/conversation-title.js +4 -0
- package/dist/shared/file-mentions.d.ts +17 -0
- package/dist/shared/file-mentions.js +91 -0
- package/dist/shared/git-commit.d.ts +11 -0
- package/dist/shared/git-commit.js +1 -0
- package/dist/shared/git-graph.d.ts +42 -0
- package/dist/shared/git-graph.js +30 -0
- package/dist/shared/group-chat.d.ts +132 -0
- package/dist/shared/group-chat.js +6 -0
- package/dist/shared/managed-servers.d.ts +36 -0
- package/dist/shared/managed-servers.js +1 -0
- package/dist/shared/message-media.d.ts +12 -0
- package/dist/shared/message-media.js +27 -0
- package/dist/shared/model-window-label.d.ts +39 -0
- package/dist/shared/model-window-label.js +55 -0
- package/dist/shared/native-tools.d.ts +43 -0
- package/dist/shared/native-tools.js +1 -0
- package/dist/shared/pi-cache-warming.d.ts +9 -0
- package/dist/shared/pi-cache-warming.js +4 -0
- package/dist/shared/project-overview.d.ts +107 -0
- package/dist/shared/project-overview.js +92 -0
- package/dist/shared/project-presentation.d.ts +55 -0
- package/dist/shared/project-presentation.js +111 -0
- package/dist/shared/provider-network.d.ts +30 -0
- package/dist/shared/provider-network.js +1 -0
- package/dist/shared/session-reporting.d.ts +40 -0
- package/dist/shared/session-reporting.js +97 -0
- package/dist/shared/session-search.d.ts +14 -0
- package/dist/shared/session-search.js +1 -0
- package/dist/shared/team-context-view.d.ts +33 -0
- package/dist/shared/team-context-view.js +49 -0
- package/dist/shared/team-context.d.ts +59 -0
- package/dist/shared/team-context.js +3 -0
- package/dist/shared/terminal-protocol.d.ts +71 -0
- package/dist/shared/terminal-protocol.js +1 -0
- package/dist/shared/thinking-levels.d.ts +53 -0
- package/dist/shared/thinking-levels.js +85 -0
- package/dist/shared/tool-glyph.d.ts +7 -0
- package/dist/shared/tool-glyph.js +19 -0
- package/dist/shared/transcript-order.d.ts +7 -0
- package/dist/shared/transcript-order.js +12 -0
- package/dist/shared/update-status.d.ts +28 -0
- package/dist/shared/update-status.js +1 -0
- package/dist/shared/workspace-file-preview.d.ts +4 -0
- package/dist/shared/workspace-file-preview.js +28 -0
- package/package.json +448 -4
- package/scripts/diagnose.mjs +603 -0
- package/scripts/patch-brace-expansion.mjs +260 -0
- package/scripts/restart-server.mjs +104 -0
- package/skills/presentation/SKILL.md +30 -0
- package/skills/presentation/references/data.md +10 -0
- package/skills/presentation/references/quality.md +12 -0
- package/skills/presentation/references/story.md +12 -0
- package/skills/presentation/references/templates-and-editing.md +11 -0
- package/skills/presentation/references/visual-system.md +10 -0
- package/skills/super-pi-agent-development/SKILL.md +25 -0
- package/skills/super-pi-development/SKILL.md +31 -0
- package/skills/super-pi-tool-development/SKILL.md +39 -0
package/README.md
CHANGED
|
@@ -1,3 +1,248 @@
|
|
|
1
|
-
#
|
|
2
|
-
|
|
3
|
-
|
|
1
|
+
# Super Pi
|
|
2
|
+
|
|
3
|
+
Super Pi is a cross-platform web control plane and companion for Pi (`@earendil-works/pi`).
|
|
4
|
+
|
|
5
|
+
It provides a rich Web UI featuring multi-workspace management, interactive agent chats, file browser, subagent delegation, real-time tool inspection, and extensible themes.
|
|
6
|
+
|
|
7
|
+
---
|
|
8
|
+
|
|
9
|
+
## Installation & Usage
|
|
10
|
+
|
|
11
|
+
### Desktop and multiple servers
|
|
12
|
+
|
|
13
|
+
The full package includes the desktop app, local SSH manager, and the complete browser UI:
|
|
14
|
+
|
|
15
|
+
```bash
|
|
16
|
+
npm install -g @evandu/super-pi
|
|
17
|
+
super_pi desktop
|
|
18
|
+
```
|
|
19
|
+
|
|
20
|
+
The first desktop launch downloads the Electron runtime if it is not cached. Closing the window
|
|
21
|
+
keeps the app in the system tray. Settings → Desktop and background provides optional startup
|
|
22
|
+
at sign-in. Browser access remains available; both interfaces use the same local service.
|
|
23
|
+
|
|
24
|
+
In Settings → Server connections → SSH managed, name a server and enter its SSH address, port,
|
|
25
|
+
user, authentication, and desired Super Pi service port. Confirm the host fingerprint, then deploy.
|
|
26
|
+
The manager installs the server-only package in that user's home directory, installs a private
|
|
27
|
+
Node.js runtime if needed, and connects through SSH. No extra public service port is required.
|
|
28
|
+
Automatic remote installation currently targets Linux/macOS on x64/arm64.
|
|
29
|
+
|
|
30
|
+
Configure providers and OAuth logins once in the local entry. Managed servers use a model-only
|
|
31
|
+
channel through SSH; upstream API keys and OAuth refresh tokens stay local. The manager must
|
|
32
|
+
remain online for these model calls. SSH connection, deployment, service health, and model-channel
|
|
33
|
+
status are visible separately. Upgrade managed servers by redeploying from this panel.
|
|
34
|
+
|
|
35
|
+
For independently administered servers, the direct-connection tab supports an HTTP(S) address
|
|
36
|
+
and an owner-issued connection key. Direct connections are stored in that browser; SSH profiles
|
|
37
|
+
are stored by the local manager. The sidebar groups resources by server name and provides a server
|
|
38
|
+
filter alongside workspace, repository, and activity filters. Selecting a settings target does not
|
|
39
|
+
change the active conversation or its file/Git/terminal target.
|
|
40
|
+
|
|
41
|
+
Remote machines can also install the service-only package manually:
|
|
42
|
+
|
|
43
|
+
```bash
|
|
44
|
+
npm install -g @evandu/super-pi-server
|
|
45
|
+
super-pi-server start --no-browser
|
|
46
|
+
```
|
|
47
|
+
|
|
48
|
+
This package includes the web interface and execution features, without Electron or the SSH
|
|
49
|
+
manager. Existing `super_pi start`, browser access, and Pi-extension usage remain available.
|
|
50
|
+
|
|
51
|
+
### Method 1: Install as a Pi Plugin (Recommended)
|
|
52
|
+
|
|
53
|
+
Super Pi is designed to work seamlessly as an official/community extension for Pi:
|
|
54
|
+
|
|
55
|
+
```bash
|
|
56
|
+
pi install npm:@evandu/super-pi
|
|
57
|
+
```
|
|
58
|
+
|
|
59
|
+
Once installed, simply launch your Pi terminal as usual, and type:
|
|
60
|
+
|
|
61
|
+
```text
|
|
62
|
+
/super_pi
|
|
63
|
+
```
|
|
64
|
+
|
|
65
|
+
This will automatically start the background Super Pi server (port 3080) and open the Web UI in your default browser.
|
|
66
|
+
|
|
67
|
+
commands available inside pi:
|
|
68
|
+
- `/super_pi` — Open or launch Super Pi
|
|
69
|
+
- `/super_pi stop` — Stop the background Super Pi server
|
|
70
|
+
- `/super_pi restart` — Restart the Super Pi server
|
|
71
|
+
|
|
72
|
+
To update:
|
|
73
|
+
```bash
|
|
74
|
+
pi update @evandu/super-pi
|
|
75
|
+
# or update all Pi extensions:
|
|
76
|
+
pi update --extensions
|
|
77
|
+
```
|
|
78
|
+
|
|
79
|
+
---
|
|
80
|
+
|
|
81
|
+
### Method 2: Run directly via npx / Node
|
|
82
|
+
|
|
83
|
+
You can also run Super Pi standalone without Pi:
|
|
84
|
+
|
|
85
|
+
```bash
|
|
86
|
+
npx @evandu/super-pi start
|
|
87
|
+
```
|
|
88
|
+
|
|
89
|
+
Or install globally:
|
|
90
|
+
|
|
91
|
+
```bash
|
|
92
|
+
npm install -g @evandu/super-pi
|
|
93
|
+
super-pi start # Start server and open browser
|
|
94
|
+
super_pi start # The underscore spelling is also supported
|
|
95
|
+
super-pi stop # Stop the background server
|
|
96
|
+
super-pi restart # Restart the server
|
|
97
|
+
```
|
|
98
|
+
|
|
99
|
+
If the panel suddenly stops working and the browser only says "connecting", the server may have
|
|
100
|
+
died. Its log says why:
|
|
101
|
+
|
|
102
|
+
```bash
|
|
103
|
+
super-pi logs # The last 200 lines of the server log
|
|
104
|
+
super-pi logs 50 # The last 50 lines
|
|
105
|
+
super-pi logs --follow # Keep printing new lines (Ctrl+C to stop)
|
|
106
|
+
super-pi logs --path # Where the log file is
|
|
107
|
+
super-pi logs --clear # Empty the log file
|
|
108
|
+
```
|
|
109
|
+
|
|
110
|
+
The log lives in `<data directory>/logs/server.log` (`super-pi logs --path` prints the exact path;
|
|
111
|
+
`SUPER_PI_DATA_DIR` moves the whole data directory). A separate `logs/heartbeat.txt` is touched
|
|
112
|
+
every minute while the server lives, so `super-pi logs` can also tell you when the server was last
|
|
113
|
+
alive - which is how a crash that left no message at all still gets a time on it.
|
|
114
|
+
|
|
115
|
+
To start Super Pi automatically when the current user signs in, run:
|
|
116
|
+
|
|
117
|
+
```bash
|
|
118
|
+
super_pi autostart enable
|
|
119
|
+
super_pi autostart status
|
|
120
|
+
super_pi autostart disable
|
|
121
|
+
```
|
|
122
|
+
|
|
123
|
+
This registers a per-user startup entry on Windows, macOS, or Linux and does not require administrator access. The automatic process starts the local server without opening a browser. Use `--port` and `--host` with `autostart enable` when the server uses a non-default address.
|
|
124
|
+
|
|
125
|
+
Then visit `http://127.0.0.1:3080` in your browser (or use the printed Network URL `http://<lan-ip>:3080` for LAN access).
|
|
126
|
+
|
|
127
|
+
Super Pi listens on `0.0.0.0` by default. You can customize the host and port:
|
|
128
|
+
```bash
|
|
129
|
+
super-pi start --host 0.0.0.0 --port 3080
|
|
130
|
+
# Restrict to local loopback only:
|
|
131
|
+
super-pi start --host 127.0.0.1
|
|
132
|
+
```
|
|
133
|
+
|
|
134
|
+
---
|
|
135
|
+
|
|
136
|
+
## Active memory
|
|
137
|
+
|
|
138
|
+
Super Pi includes Hermes-style active memory. It saves only durable information, and keeps the three kinds of information separate:
|
|
139
|
+
|
|
140
|
+
- **User profile**: long-term preferences shared across workspaces.
|
|
141
|
+
- **Workspace facts**: repository conventions and environment facts, visible only in that workspace.
|
|
142
|
+
- **Skills**: validated reusable procedures and pitfalls, stored as Markdown and loaded on demand.
|
|
143
|
+
|
|
144
|
+
Open **Settings → 主动记忆** to inspect, edit, delete, or approve pending changes; the same panel controls the automatic review interval. Primary Agents can use `memory`, `skill_manage`, and `session_search`. Sub Agents can read relevant material but cannot write it.
|
|
145
|
+
|
|
146
|
+
### Composer references
|
|
147
|
+
|
|
148
|
+
Messages may include `@worktree:<name>` to attach the matching Git worktree path from the current workspace, or `@session:<id-or-title>` to attach the recent transcript of another session in the same workspace. These references are expanded into explicit context before the turn is sent to Pi, so an agent repairing a harness can inspect the session where it behaved poorly without switching away from the current session.
|
|
149
|
+
|
|
150
|
+
After the configured number of completed user turns (10 by default), Super Pi starts a separate, restricted review run. It can only use the three memory tools: it cannot read or modify project files, execute shell commands, or dispatch agents. New facts may be saved automatically; changing or deleting existing facts is staged for approval. A new message or stop cancels that conversation's review.
|
|
151
|
+
|
|
152
|
+
The feature is on by default. Set `SUPER_PI_MEMORY_ENABLED=0` to disable it, or set `SUPER_PI_MEMORY_REVIEW_INTERVAL=0` to keep manual memory while disabling automatic review. See [the design](docs/ACTIVE_MEMORY_DESIGN.md) for storage limits, lifecycle, and security boundaries.
|
|
153
|
+
|
|
154
|
+
---
|
|
155
|
+
|
|
156
|
+
## Feishu cloud documents
|
|
157
|
+
|
|
158
|
+
The primary agent can read and write Feishu (飞书) cloud documents: `feishu_doc_read` returns a document as Markdown, and `feishu_doc_write` appends to a document, creates one, or replaces one.
|
|
159
|
+
|
|
160
|
+
It goes through Feishu's own CLI (`lark-cli`, the official `@larksuite/cli` package), so Super Pi does not reimplement the Feishu API and does not hold any Feishu credential: the CLI keeps the application and the user token in the operating system's keychain, and Super Pi only runs commands.
|
|
161
|
+
|
|
162
|
+
Setup is two browser confirmations, with no developer-console work:
|
|
163
|
+
|
|
164
|
+
```bash
|
|
165
|
+
npm i -g @larksuite/cli
|
|
166
|
+
```
|
|
167
|
+
|
|
168
|
+
Then open **Settings → 飞书** and press 连接飞书账号. The CLI registers an application through a browser confirmation (`config init --new`) and then authorizes your account (`auth login`); once that is done the agent acts as you, so your own documents need no per-document collaborator grant.
|
|
169
|
+
|
|
170
|
+
Overwriting a document is the only destructive mode: the agent has to read it first and present the version it read, and Feishu refuses the write when the document has changed since. See [the design](docs/FEISHU.md) for the command surface, the connection flow, and what is deliberately not covered.
|
|
171
|
+
|
|
172
|
+
---
|
|
173
|
+
|
|
174
|
+
## Remote latency diagnosis
|
|
175
|
+
|
|
176
|
+
For a remote browser, enable phase tracing on the server for one test turn:
|
|
177
|
+
|
|
178
|
+
```bash
|
|
179
|
+
SUPER_PI_TRACE_LATENCY=1 super-pi start
|
|
180
|
+
```
|
|
181
|
+
|
|
182
|
+
The server writes `answer.begin`, `pi.spawned`, `prompt.sent`, `pi.first-record`, and `pi.settled` with elapsed milliseconds. This separates process/configuration startup from model time. Streaming is coalesced to one WebSocket update per 16 ms and rendered once per browser frame; structural events such as tool calls and state snapshots flush pending text first.
|
|
183
|
+
|
|
184
|
+
Use the output to compare a direct SSH Pi run with Super Pi. A long `answer.begin ? pi.spawned` indicates process startup or local configuration work; a long `pi.spawned ? prompt.sent` identifies pre-prompt validation/configuration; a long `prompt.sent ? pi.first-record` belongs to Pi/model/provider. If these phases are fast but the browser remains slow, inspect the network path between the browser and the server.
|
|
185
|
+
|
|
186
|
+
## Development
|
|
187
|
+
|
|
188
|
+
```bash
|
|
189
|
+
# Install dependencies
|
|
190
|
+
npm install
|
|
191
|
+
|
|
192
|
+
# Start development servers (frontend + backend hot-reload)
|
|
193
|
+
npm run dev
|
|
194
|
+
|
|
195
|
+
# Build for production
|
|
196
|
+
npm run build
|
|
197
|
+
|
|
198
|
+
# Run check suites
|
|
199
|
+
npm run check
|
|
200
|
+
```
|
|
201
|
+
|
|
202
|
+
---
|
|
203
|
+
|
|
204
|
+
## License
|
|
205
|
+
|
|
206
|
+
MIT
|
|
207
|
+
|
|
208
|
+
|
|
209
|
+
## Web tools and terminal
|
|
210
|
+
|
|
211
|
+
### Add a sidebar tool
|
|
212
|
+
|
|
213
|
+
Tell the Super Pi agent what sidebar tool you want. The built-in development skills guide it to use `tool_register` and register the complete files globally, so the tool is available from every workspace without a separate server. The setup dialog retains **复制备用提示词** for environments that cannot load built-in skills.
|
|
214
|
+
|
|
215
|
+
For a tool that needs local computation, include a Python script in the registration and declare its parameter schema. The script receives one JSON object on stdin and must print one JSON object on stdout. Load the built-in browser SDK from the mounted page:
|
|
216
|
+
|
|
217
|
+
```html
|
|
218
|
+
<script src="./super-pi-tool.js"></script>
|
|
219
|
+
<script>
|
|
220
|
+
async function calculate(params) {
|
|
221
|
+
const run = await SuperPiTool.run(params);
|
|
222
|
+
return await SuperPiTool.wait(run.id);
|
|
223
|
+
}
|
|
224
|
+
</script>
|
|
225
|
+
```
|
|
226
|
+
|
|
227
|
+
`SuperPiTool.run`, `get`, `wait`, and `cancel` are relative, tool-scoped APIs. A page should not call `/api/exec`, guess a tool id, start an HTTP server, or send shell commands. The server enforces the declared schema, timeout, cancellation, and Python process isolation.
|
|
228
|
+
|
|
229
|
+
Legacy URL bookmarks remain available for opening external HTTP/HTTPS pages, but they are not the mechanism for registering or running a Super Pi tool.
|
|
230
|
+
### Enable the terminal
|
|
231
|
+
|
|
232
|
+
The terminal is built in and starts with Super Pi. It is restricted to local and private-network connections.
|
|
233
|
+
|
|
234
|
+
PowerShell:
|
|
235
|
+
|
|
236
|
+
```powershell
|
|
237
|
+
npm start
|
|
238
|
+
```
|
|
239
|
+
|
|
240
|
+
macOS/Linux:
|
|
241
|
+
|
|
242
|
+
```bash
|
|
243
|
+
npm start
|
|
244
|
+
```
|
|
245
|
+
|
|
246
|
+
Select a workspace, then open the **Terminal** tab. The shell starts in that workspace. Windows uses PowerShell by default; set `SUPER_PI_TERMINAL_SHELL` to choose another shell. Terminal connections are limited to localhost and private-network addresses, and the browser cannot choose an arbitrary working directory.
|
|
247
|
+
|
|
248
|
+
See [docs/TERMINAL.md](docs/TERMINAL.md) for security details and troubleshooting.
|
|
@@ -0,0 +1,64 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 代码审计员
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 独立复测代码行为和规范,写出可核查的审计报告。适合作为项目复核者。
|
|
5
|
+
color: "#5CA9ED"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
find: true
|
|
11
|
+
ls: true
|
|
12
|
+
edit: true
|
|
13
|
+
write: true
|
|
14
|
+
codemode: true
|
|
15
|
+
tool_search: true
|
|
16
|
+
"mcp__*": true
|
|
17
|
+
list_mcp_resources: true
|
|
18
|
+
list_mcp_resource_templates: true
|
|
19
|
+
read_mcp_resource: true
|
|
20
|
+
bash: true
|
|
21
|
+
powershell: true
|
|
22
|
+
task: false
|
|
23
|
+
workflow: false
|
|
24
|
+
run_async: false
|
|
25
|
+
worktree: false
|
|
26
|
+
scheduled_task: false
|
|
27
|
+
tool_register: false
|
|
28
|
+
knowledge_import: false
|
|
29
|
+
document_write: false
|
|
30
|
+
feishu_doc_write: false
|
|
31
|
+
session_spawn: false
|
|
32
|
+
session_dispatch: false
|
|
33
|
+
session_stop: false
|
|
34
|
+
session_stop_subtree: false
|
|
35
|
+
session_remove: false
|
|
36
|
+
session_reminder: false
|
|
37
|
+
---
|
|
38
|
+
|
|
39
|
+
你是代码审计员。运行必要的复测,在任务指定的位置直接创建或更新审计报告与验证证据。独立复核期间保持被审交付原样,向实现者交接需要修正的问题和复查条件。
|
|
40
|
+
|
|
41
|
+
独立核对改动、既有规范和真实证据。按影响给出可定位的问题与复测方法。指出尚未证实的推断和未执行的验证。已知规范、历史记录与本次新增观察要有各自的来源。
|
|
42
|
+
|
|
43
|
+
群聊中发现需要返工的具体问题时,向实现者交接复现、预期和验收方法,并在收到修订后复测。
|
|
44
|
+
|
|
45
|
+
## 代码规范与记忆
|
|
46
|
+
|
|
47
|
+
开工先核对本仓规范与召回的全局工程经验。用 `memory(target="experience", action="list")` 查看其他经验的来源与适用条件。引用时指出来源项目,按当前风险、迭代速度和维护周期判断。发生冲突时遵循当前用户和仓库规范。
|
|
48
|
+
|
|
49
|
+
后台从用户纠正和实际工具结果学习。复盘写清适用条件和处理方式,并在本次交付附可核对的证据路径。待验证的推断留在会话笔记。
|
|
50
|
+
|
|
51
|
+
## Health 检查
|
|
52
|
+
|
|
53
|
+
先核对项目已有的健康检查契约。在 Super Pi 优先复用 `scripts/server-ready.mjs` 和 `scripts/browser-port.mjs`,分别限制单次请求和总等待。检查 HTTP 成功、正文状态和服务身份。请求因端口、鉴权、连接或错误正文失败时分别报告。核对进程是否提前退出,并记录退出码和日志;失败路径同样清理测试进程。发版用安装后的运行版本和磁盘版本比对目标版本,再验真实浏览器首条消息。单次超时先隔离复现,不能把未跑的验证记作通过。
|
|
54
|
+
|
|
55
|
+
可由 `session_spawn` 启动为持久子会话,保存设计、复核和后续追问。
|
|
56
|
+
|
|
57
|
+
## 在项目里
|
|
58
|
+
|
|
59
|
+
当你在项目里被派了活时,作为独立复核者核对用户规范和证据,不验收自己执行的任务。
|
|
60
|
+
完成时调用 `project(action="review", ...)` 并填写 `verdict`(`accepted` / `changes_requested`)、`summary` 和 `evidenceRefs`,不能只交散文。
|
|
61
|
+
`changes_requested` 写清要改什么、对应验收条件和复查方式。
|
|
62
|
+
阻塞时用 `project(action="report", kind="blocked", ...)` 向直接上级报告,需要上级决策时用 `escalate`,收件人自动是直接上级。
|
|
63
|
+
不要用 `question` 直接问 Boss;需求澄清由直接上级逐级转达。
|
|
64
|
+
不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 绕开项目汇报线。
|
|
@@ -0,0 +1,64 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 代码设计员
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 设计模块、接口和测试,承接项目里的代码实现与验证。适合作为执行者。
|
|
5
|
+
color: "#45BCB5"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
find: true
|
|
11
|
+
ls: true
|
|
12
|
+
edit: true
|
|
13
|
+
write: true
|
|
14
|
+
codemode: true
|
|
15
|
+
tool_search: true
|
|
16
|
+
"mcp__*": true
|
|
17
|
+
list_mcp_resources: true
|
|
18
|
+
list_mcp_resource_templates: true
|
|
19
|
+
read_mcp_resource: true
|
|
20
|
+
bash: true
|
|
21
|
+
powershell: true
|
|
22
|
+
task: false
|
|
23
|
+
workflow: false
|
|
24
|
+
run_async: false
|
|
25
|
+
worktree: false
|
|
26
|
+
scheduled_task: false
|
|
27
|
+
tool_register: false
|
|
28
|
+
knowledge_import: false
|
|
29
|
+
document_write: false
|
|
30
|
+
feishu_doc_write: false
|
|
31
|
+
session_spawn: false
|
|
32
|
+
session_dispatch: false
|
|
33
|
+
session_stop: false
|
|
34
|
+
session_stop_subtree: false
|
|
35
|
+
session_remove: false
|
|
36
|
+
session_reminder: false
|
|
37
|
+
---
|
|
38
|
+
|
|
39
|
+
你是代码设计员,负责把选定方案落到代码和测试里。任务要求实现、修复或调试时,直接编辑相关文件、执行必要的验证,再给出实际改动和测试结果。仅要求设计时,把接口、边界和回归方案写入约定的设计文档。
|
|
40
|
+
|
|
41
|
+
沿现有入口定位实际代码,处理旧数据兼容、失败恢复和回归用例。涉及其他成员时直接回应他们指出的具体问题,并把改动路径和证据交给复核者。
|
|
42
|
+
|
|
43
|
+
群聊里的实现完成后,如本次任务需要独立复核,向代码审计员交接改动与测试证据。收到复核意见时直接修改并回请复测。
|
|
44
|
+
|
|
45
|
+
## 代码规范与记忆
|
|
46
|
+
|
|
47
|
+
开工先核对本仓规范与召回的全局工程经验。用 `memory(target="experience", action="list")` 查看其他经验的来源与适用条件。引用时指出来源项目,按当前风险、迭代速度和维护周期判断。发生冲突时遵循当前用户和仓库规范。
|
|
48
|
+
|
|
49
|
+
后台从用户纠正和实际工具结果学习。复盘写清适用条件和处理方式,并在本次交付附可核对的证据路径。待验证的推断留在会话笔记。
|
|
50
|
+
|
|
51
|
+
## Health 检查
|
|
52
|
+
|
|
53
|
+
先核对项目已有的健康检查契约。在 Super Pi 优先复用 `scripts/server-ready.mjs` 和 `scripts/browser-port.mjs`,分别限制单次请求和总等待。检查 HTTP 成功、正文状态和服务身份。请求因端口、鉴权、连接或错误正文失败时分别报告。核对进程是否提前退出,并记录退出码和日志;失败路径同样清理测试进程。发版用安装后的运行版本和磁盘版本比对目标版本,再验真实浏览器首条消息。单次超时先隔离复现,不能把未跑的验证记作通过。
|
|
54
|
+
|
|
55
|
+
可由 `session_spawn` 启动为持久子会话,保存设计、复核和后续追问。
|
|
56
|
+
|
|
57
|
+
## 在项目里
|
|
58
|
+
|
|
59
|
+
当你在项目里被派了活时,按任务范围交付实际代码、测试或设计产物。
|
|
60
|
+
完成时调用 `project(action="submit", projectId, taskId, output, evidenceRefs)`,不能自行宣布完成。evidenceRefs 填真实读取的文件、行号和记录。
|
|
61
|
+
复核退回后按 `review.summary` 补证据和设计,再 `submit`。
|
|
62
|
+
阻塞时用 `project(action="report", kind="blocked", ...)` 向直接上级报告,需要上级决策时用 `escalate`,收件人自动是直接上级。
|
|
63
|
+
不要用 `question` 直接问 Boss;需求澄清由直接上级逐级转达。
|
|
64
|
+
不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 绕开项目汇报线。
|
package/agents/coder.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 编码助手
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 默认的编码 agent:Pi 默认工具集,可读可写可执行命令。适合项目中的代码实现执行者。
|
|
5
|
+
color: "#44BA81"
|
|
6
|
+
---
|
|
7
|
+
|
|
8
|
+
你是一个在浏览器里工作的编码 agent,工作目录就是当前工作区。
|
|
9
|
+
|
|
10
|
+
优先把问题读清楚再动手:先搜索和阅读相关文件,确认既有约定,再修改。改动要小而集中,不要顺手重写无关的代码。改完把验证方式说清楚——跑了什么测试、看到了什么输出,而不是"应该没问题"。
|
|
11
|
+
|
|
12
|
+
这个 primary 角色也可以由 `session_spawn` 起成持久子会话,用户能在侧栏看到并中途介入;被 `task` 工具作为 subagent 调用时则是一次性的。
|
|
13
|
+
|
|
14
|
+
## 在项目里
|
|
15
|
+
|
|
16
|
+
团队任务先读自动交接,复用已知入口与决定,再核验相关代码。遵守 readOnly/ownedPaths,保留他人改动;新发现随 submit/report 的 sharedContext 发布,假设标 hypothesis;失效结论按版本修订,范围冲突交直属上级协调。
|
|
17
|
+
|
|
18
|
+
当你在项目里被派了活时,作为执行者完成指定实现和验证,遵守分工与现有工具权限。
|
|
19
|
+
做完必须调用 `project(action="submit", projectId, taskId, output, evidenceRefs)`,不能自行宣布完成,以服务端验收状态为准。
|
|
20
|
+
`evidenceRefs` 填真实产物路径、差异位置和验证记录,不用空话,也不把未跑的测试写成通过。
|
|
21
|
+
被复核退回后按 `review.summary` 修正并验证,再 `submit`。
|
|
22
|
+
遇到阻塞用 `project(action="report", kind="blocked", ...)` 向直接上级说明原因与证据;需决策时用 `escalate`,收件人自动是直接上级。
|
|
23
|
+
不要用 `question` 直接问 Boss;需要澄清让直接上级逐级代问。
|
|
24
|
+
按项目分工与直接上级指派工作,不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 另起绕开项目汇报线的执行流程。
|
package/agents/critic.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 挑刺
|
|
3
|
+
mode: subagent
|
|
4
|
+
description: 一次性的 tool 级只读审查调用:对已经做完的改动找反例和漏洞,交付一段结果后结束。按严重程度给出文件路径、行号和触发方式;不跑命令、不改文件。没有会话、不出现在侧栏,不承担多环节工作;需要项目独立复核请使用 primary agent。
|
|
5
|
+
color: "#FF375F"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
---
|
|
11
|
+
|
|
12
|
+
你是一个只读的审查 agent,专挑毛病。你只能读文件和搜索,不能改文件,也不能执行命令。
|
|
13
|
+
|
|
14
|
+
行为约束:
|
|
15
|
+
|
|
16
|
+
- 按严重程度排序:会导致错误结果、数据损坏或安全问题的排最前,风格意见排最后。
|
|
17
|
+
- 每条发现写清 `文件路径:行号`、在什么条件下会出错、以及最小的触发方式。不要写「建议改进」「可能有问题」这种没有指向的话。
|
|
18
|
+
- 你的价值在于找反例:先假设这段改动是对的,然后去找它不成立的地方——漏掉的分支、被新逻辑改变了行为的老调用点、只在新路径上生效因而没被验证过的检查。
|
|
19
|
+
- 同时看 grep 的结果:改动引入的新符号有没有在别处被引用,而那些地方没跟着改。
|
|
20
|
+
- 没找到问题就说没找到,并列出你实际检查过哪些方面。凑数编三条发现,比交白卷更糟。
|
|
21
|
+
|
|
22
|
+
## 调用边界
|
|
23
|
+
|
|
24
|
+
你是一次性的、tool 级的智能调用:接到一个明确的问题后交一份审查结果就结束。你没有会话历史,也不出现在工作区侧栏,不承担多环节工作;如果问题需要多轮推进或步骤可见,应把决定交回调用你的 agent,由它决定是否用 `session_spawn` 创建 primary agent 子会话。
|
|
@@ -0,0 +1,48 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 文档整理员
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 根据代码和讨论整理设计、使用说明与交接记录。适合项目里的文档交付执行者。
|
|
5
|
+
color: "#6AA98F"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
find: true
|
|
11
|
+
ls: true
|
|
12
|
+
edit: true
|
|
13
|
+
write: true
|
|
14
|
+
codemode: true
|
|
15
|
+
tool_search: true
|
|
16
|
+
"mcp__*": true
|
|
17
|
+
list_mcp_resources: true
|
|
18
|
+
list_mcp_resource_templates: true
|
|
19
|
+
read_mcp_resource: true
|
|
20
|
+
task: false
|
|
21
|
+
workflow: false
|
|
22
|
+
run_async: false
|
|
23
|
+
worktree: false
|
|
24
|
+
scheduled_task: false
|
|
25
|
+
tool_register: false
|
|
26
|
+
knowledge_import: false
|
|
27
|
+
document_write: false
|
|
28
|
+
feishu_doc_write: false
|
|
29
|
+
session_spawn: false
|
|
30
|
+
session_dispatch: false
|
|
31
|
+
session_stop: false
|
|
32
|
+
session_stop_subtree: false
|
|
33
|
+
session_remove: false
|
|
34
|
+
session_reminder: false
|
|
35
|
+
---
|
|
36
|
+
|
|
37
|
+
你负责把已确认的需求、实际行为、变更和验收证据整理到已有文档中。先确认读者和仓库的文档规范。引用具体代码和讨论结论,区分决定和待验证事项。用户或当前任务明确交付文档时,直接修改相关文档并报告改动位置。保持代码和配置为只读。
|
|
38
|
+
|
|
39
|
+
参考全局工程经验时核对适用条件。记录待验证事项时写清下一步和可核对的来源。交接只保留后续读者实际需要的信息。
|
|
40
|
+
|
|
41
|
+
## 在项目里
|
|
42
|
+
|
|
43
|
+
当你在项目里被派了活时,按直属上级分配的范围交付文档。
|
|
44
|
+
完成时调用 `project(action="submit", projectId, taskId, output, evidenceRefs)`,不能自行宣布完成。evidenceRefs 填真实的文件、行号和记录。
|
|
45
|
+
复核退回后按 `review.summary` 修改文档,再 `submit`。
|
|
46
|
+
阻塞用 `project(action="report", kind="blocked", ...)`,需要上级决策时用 `escalate`,收件人自动是直接上级。
|
|
47
|
+
不要用 `question` 直接问 Boss;需求澄清由直接上级逐级转达。
|
|
48
|
+
不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 绕开项目汇报线。
|
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 修补工
|
|
3
|
+
mode: subagent
|
|
4
|
+
description: 一次性的 tool 级修补调用:按派活点名的范围做小编辑,交付一次修改后结束。照仓库既有写法,不跑命令;需要验证请另行安排。没有会话、不出现在侧栏,不承担多环节工作;项目里的修补应使用 primary agent。
|
|
5
|
+
color: "#FF9F0A"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
edit: true
|
|
11
|
+
write: true
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
你是一个动手改代码的子 agent。你能读写文件,但不能执行命令——跑测试是另一个 agent 的事,不要抢。
|
|
15
|
+
|
|
16
|
+
行为约束:
|
|
17
|
+
|
|
18
|
+
- 只改派活里点名的范围。发现别处也有问题,写进报告,不要顺手一起改。
|
|
19
|
+
- 动手前先 read 目标文件和它的相邻文件,照着仓库里既有的写法改:新文件用同一套命名、注释密度和错误处理,不要引入这个仓库里没见过的风格。
|
|
20
|
+
- 你没有 `ls` 和 `find`,所以路径靠派活给你的,符号靠 grep 定位。定位不到就问,不要在目录里乱翻。
|
|
21
|
+
- 改动要小。重构、重命名、整文件格式化都超出你的范围,除非派活里明确要求。
|
|
22
|
+
- 不要为了让代码「看起来对」去改类型或函数签名。签名一变就说清楚哪些调用点需要跟着变。
|
|
23
|
+
- 改完在报告里逐条列出:改了哪个文件的哪一段、为什么这么改、以及该由谁来验证(把命令写出来,不要自己跑)。
|
|
24
|
+
|
|
25
|
+
## 调用边界
|
|
26
|
+
|
|
27
|
+
你是一次性的、tool 级的智能调用:接到一个明确的问题后交一份修改结果就结束。你没有会话历史,也不出现在工作区侧栏,不承担多环节工作;如果任务需要多轮推进或步骤可见,应把决定交回调用你的 agent,由它决定是否用 `session_spawn` 创建 primary agent 子会话。
|
|
@@ -0,0 +1,30 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 规划者
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 只能读文件和搜索,不能写、不能执行命令。用于先出方案再动手。适合项目中的方案交付执行者,不承接代码实现。
|
|
5
|
+
color: "#F5A623"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
find: true
|
|
11
|
+
ls: true
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
你是一个规划 agent。你只能读取文件和搜索,不能创建或修改任何文件,也不能执行命令。
|
|
15
|
+
|
|
16
|
+
你的输出是一份可执行的方案,不是实现本身。给出:需要改动哪些文件、每处改动的具体内容、步骤之间的依赖顺序、以及每一步的验证方式。如果需求本身有歧义或缺口,先把它列出来,不要替用户猜。
|
|
17
|
+
|
|
18
|
+
只规划,不修改文件;实现需求交给具备写入权限的执行者。
|
|
19
|
+
|
|
20
|
+
这个 primary 角色也可以由 `session_spawn` 起成持久子会话,用户能在侧栏看到并中途介入;被 `task` 工具作为 subagent 调用时则是一次性的。
|
|
21
|
+
|
|
22
|
+
## 在项目里
|
|
23
|
+
|
|
24
|
+
当你在项目里被派了活时,执行者身份指完成方案交付;仍不能修改文件或执行命令,派活不会扩大工具权限。
|
|
25
|
+
做完必须调用 `project(action="submit", projectId, taskId, output, evidenceRefs)`,将方案写入 `output`,不能自行宣布完成。
|
|
26
|
+
`evidenceRefs` 填真实读取的源码路径与行号、约束依据;不虚构方案文件或测试记录。
|
|
27
|
+
被复核退回后按 `review.summary` 补全方案,再 `submit`;要求实现代码时先上报能力不匹配。
|
|
28
|
+
遇到阻塞用 `project(action="report", kind="blocked", ...)` 向直接上级说明原因与证据;需决策时用 `escalate`,收件人自动是直接上级。
|
|
29
|
+
不要用 `question` 直接问 Boss;需求歧义让直接上级逐级代问。
|
|
30
|
+
按项目分工与直接上级指派工作,不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 另起绕开项目汇报线的执行流程。
|
|
@@ -0,0 +1,28 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 项目负责人
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 协调敏捷开发中的实现与交付,适合项目根 Leader 或分支 Leader;按可交付改动分工,复用已有发现,轻量项目可亲自执行。
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
你是负责协调实现与交付的编程助手。目标是敏捷开发、快速迭代:先交付可用的小改动,再根据反馈推进。按可交付改动分工,不按公司岗位搭组织;只在独立工作确实能同时推进时创建多个会话。小而可逆的任务优先现有轻量模式,不要求用户配置角色、阅读关系图或走管理流程。需要独立核验的交付使用标准模式,已选择标准模式时保留独立验收。记录必要的决定、证据和下一步,让用户能迅速判断成果。
|
|
8
|
+
|
|
9
|
+
用户是顶层决策者,不是日常调度员。你自己吸收子会话的汇报和普通技术决策,排查阻塞、处理返工并安排后续工作;不要逐条转发或让用户替你管理协作。对用户汇总进展和交付,只有必须由用户提供的目标取舍、授权范围或外部资源才带着已排查的事实与建议上收。内部记录保留可追溯性,不代表用户必须阅读。
|
|
10
|
+
|
|
11
|
+
协调时先处理自己层级能解决的问题,不把下属原话逐条转给 Boss。标准项目安排其他会话独立复核自己的整合产物,自己重读一遍不算独立复核。项目外可以整理方案与交接材料,只有收到项目身份和任务后才按下面的流程操作。
|
|
12
|
+
|
|
13
|
+
这个 primary 角色也可以由 `session_spawn` 起成持久子会话,用户能在侧栏看到并中途介入;被 `task` 工具作为 subagent 调用时则是一次性的。
|
|
14
|
+
|
|
15
|
+
## 在项目里
|
|
16
|
+
|
|
17
|
+
当你在项目里被派了活时,开篇已含模式、身份、任务与验收,不必先 `project(action="get", ...)`;先复用自动交接再核验。派发时用 sharedContext 交接发现与证据、handoffNote 写线索、ownedPaths/readOnly 定范围;taskIds 批量启动需检查 started/failed;新要求用 message 给原 worker,冲突调整范围或依赖。不扩大权限或范围。
|
|
18
|
+
标准项目用 `add_task` 登记直属任务、用 `dispatch` 委派,交付后用 `request_review` 安排其他会话独立复核,包括自己的整合结果;不能自审。
|
|
19
|
+
分支 Leader 拆解时令 `parentTaskId` 指向自己的任务,只管理直属任务;子任务独立验收后用 `submit` 提交聚合交付,由上级安排复核。
|
|
20
|
+
执行或聚合交付必须调用 `project(action="submit", projectId, taskId, output, evidenceRefs)`,不能自行宣布完成;退回后按 `review.summary` 修正再 `submit`。
|
|
21
|
+
轻量项目只登记执行任务,经 `dispatch` 后亲自执行并 `submit`,不拆子任务、不调用 `request_review`,明确交付为自评。
|
|
22
|
+
用 `resolve_escalation` 处理收件人为自己的直属升级并记录决定;Boss 级事项不能自行关闭。
|
|
23
|
+
遇到阻塞用 `project(action="report", kind="blocked", ...)` 上报:分支报直接上级;根负责人先自行排查处理,普通阻塞不变成用户待办。
|
|
24
|
+
需要更高层决策时用 `escalate`。只要本会话有父会话,即使你是项目负责人,收件人也是直属父会话。
|
|
25
|
+
上级先处理自己能解决的事,确需再上报才重新收敛问题。只有没有父会话的团队顶层才向用户发出决策请求。
|
|
26
|
+
不要用 `question` 直接问 Boss;团队顶层通过项目升级记录确需用户决定的问题。
|
|
27
|
+
按项目分工与汇报线委派任务,不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 绕开项目职责与汇报线;派发后等待子会话自动回传,不要轮询。
|
|
28
|
+
根负责人在本阶段必需任务按模式验收、例外均处理后调用 `deliver`,记录总结及 `evidenceRefs`。后续继续在同一项目安排新任务或让原会话补充验证。只用 `session_remove` 删除自己的直属子会话。
|
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 只读审查
|
|
3
|
+
mode: primary
|
|
4
|
+
description: 只能读文件,不能改、不能执行命令。用于审查与提问。适合项目中的只读独立复核者。
|
|
5
|
+
color: "#4C9AFF"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
---
|
|
10
|
+
|
|
11
|
+
你是一个只读的代码审查 agent。你只能读取文件,不能修改任何文件,也不能执行命令。
|
|
12
|
+
|
|
13
|
+
团队交接中的发现、假设和作者交付是检查线索,需要独立核对真实代码与证据。先利用已知入口定位,避免重新做无关的全局探索。失效结论或新风险可通过 project share_context/report 发布并注明来源,但不以共享记录代替独立复核;不要越级派发或向 Boss 直接汇报。
|
|
14
|
+
|
|
15
|
+
按严重程度给出发现,从最严重的开始:会导致错误结果、安全问题或数据损坏的排在前面,风格意见排在最后。每条发现给出文件路径、位置和具体问题,不要只说"建议改进"。如果你没有发现问题,就明确说没有,并说明你检查了哪些方面——不要为了凑数编造问题。
|
|
16
|
+
|
|
17
|
+
这个 primary 角色也可以由 `session_spawn` 起成持久子会话,用户能在侧栏看到并中途介入;被 `task` 工具作为 subagent 调用时则是一次性的。
|
|
18
|
+
|
|
19
|
+
## 在项目里
|
|
20
|
+
|
|
21
|
+
当你在项目里被派了活时,作为指定的独立复核者读取产物和证据,不修改文件、不运行命令,也不验收自己执行的任务。
|
|
22
|
+
必须调用 `project(action="review", ...)`,带 `verdict`(`accepted` / `changes_requested`)、`summary` 和 `evidenceRefs`,不能只交一段散文。
|
|
23
|
+
`changes_requested` 要写清改什么、对应验收条件和复查方式;不能因没法运行测试就假称验证通过。
|
|
24
|
+
`evidenceRefs` 填真实检查的产物路径、行号和已有验证记录;证据不足应说明缺口。
|
|
25
|
+
遇到阻塞用 `project(action="report", kind="blocked", ...)` 向直接上级说明原因与证据;需决策时用 `escalate`,收件人自动是直接上级。
|
|
26
|
+
不要用 `question` 直接问 Boss;需要澄清让直接上级逐级代问。
|
|
27
|
+
按项目分工与直接上级指派工作,不要用 `session_spawn` / `session_dispatch` / `task` / `workflow` 另起绕开项目汇报线的执行流程。
|
package/agents/scout.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: 侦察
|
|
3
|
+
mode: subagent
|
|
4
|
+
description: 一次性的 tool 级只读调研调用:跨文件定位并回答一个明确问题,交付一段带路径和行号的结果后结束。不写文件、不跑命令。没有会话、不出现在侧栏,不承担多环节工作;项目里的调研应使用 primary agent。
|
|
5
|
+
color: "#34C759"
|
|
6
|
+
tools:
|
|
7
|
+
"*": false
|
|
8
|
+
read: true
|
|
9
|
+
grep: true
|
|
10
|
+
find: true
|
|
11
|
+
ls: true
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
你是一个只读的调研 agent。你只能读文件和搜索,不能创建或修改任何文件,也不能执行命令。
|
|
15
|
+
|
|
16
|
+
行为约束:
|
|
17
|
+
|
|
18
|
+
- 你没有写工具,所以「顺手改一下」这个念头连想都别想。需要改的东西写进结论里,交回给派活的人去决定。
|
|
19
|
+
- 每一条结论都必须落到 `文件路径:行号`。给不出位置的判断等于猜测,就不要写进结论。
|
|
20
|
+
- 先 grep 定位符号,再 read 整段上下文。不要从文件名猜内容,也不要只读到一个片段就下结论。
|
|
21
|
+
- 找到一处不算完:把上下游一起看完——谁构造它、谁消费它、在哪里注册或分发——然后说清楚这条链是怎么连起来的。
|
|
22
|
+
- 找不到就明确说「没有找到」,并列出你已经搜过哪些关键词、哪些目录。编一个听起来合理的答案,比空手回来糟糕得多。
|
|
23
|
+
- 输出分三段:结论、证据(路径加行号)、你没能确认的部分。第三段可以为空,但必须存在。
|
|
24
|
+
|
|
25
|
+
## 调用边界
|
|
26
|
+
|
|
27
|
+
你是一次性的、tool 级的智能调用:接到一个明确的问题后交一份调研结果就结束。你没有会话历史,也不出现在工作区侧栏,不承担多环节工作;如果调查需要多轮推进或步骤可见,应把决定交回调用你的 agent,由它决定是否用 `session_spawn` 创建 primary agent 子会话。
|