Skip to content
Back to search

hello2cc-local

MarketplaceWorks with Claude CodeGitHub

Native-first Claude Code plugin for third-party models with silent Agent model injection and output styles.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/hellowind777-hello2cc/badge.svg)](/listings/hellowind777-hello2cc)

HTML

<a href="/listings/hellowind777-hello2cc"><img src="/api/listings/hellowind777-hello2cc/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Prerequisites

- Node.js 18 or later - Claude Code with plugin support - a working third-party model mapping layer such as CCSwitch, if you are not using native Claude models

Install

1. Clone the repository.

   git clone https://github.com/hellowind777/hello2cc.git
   cd hello2cc

2. Add the local marketplace entry.

   claude plugins marketplace add "<repo-path>"

Replace `<repo-path>` with your local `hello2cc` repository path.

3. Install and enable the plugin.

   claude plugins install hello2cc@hello2cc-local
   claude plugins enable hello2cc@hello2cc-local

4. Reload Claude Code.

   /reload-plugins

Verify

Run:

claude plugins list

Expected result:

- `hello2cc@hello2cc-local` is installed - the plugin status is enabled - new sessions no longer rely on a plugin-shipped `settings.json` to force `agent=hello2cc:native`

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 20 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "hello2cc-local",
  3. "owner": {
  4. "name": "hello2cc"
  5. },
  6. "metadata": {
  7. "description": "Local marketplace for the hello2cc Claude Code plugin"
  8. },
  9. "plugins": [
  10. {
  11. "name": "hello2cc",
  12. "description": "Thin Claude Code plugin for third-party models, focused on host state, protocol adaptation, and repeated-failure debounce.",
  13. "version": "0.5.11",
  14. "source": "./",
  15. "author": {
  16. "name": "hello2cc"
  17. }
  18. }
  19. ]
  20. }
README.md · README · 252 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # hello2cc
  2. [![npm version](https://img.shields.io/npm/v/hello2cc.svg)](https://www.npmjs.com/package/hello2cc)
  3. [![License](https://img.shields.io/badge/license-Apache%202.0-blue.svg)](./LICENSE)
  4. [![Publish](https://img.shields.io/github/actions/workflow/status/hellowind777/hello2cc/publish.yml?label=publish)](https://github.com/hellowind777/hello2cc/actions/workflows/publish.yml)
  5. Make third-party models inside Claude Code behave closer to a native Opus session.
  6. `hello2cc` does not replace your gateway, provider mapping, or account setup.
  7. It sits in the plugin layer and keeps third-party models closer to Claude Code's native behavior around tool choice, agent routing, task and team workflow, failure handling, and response style.
  8. **Language:** English | [简体中文](./README_CN.md)
  9. ## Overview
  10. `hello2cc` is for users who already run GPT, Kimi, DeepSeek, Gemini, Qwen, or other third-party models in Claude Code through CCSwitch or a similar mapping layer.
  11. It helps in three places:
  12. - keep capability selection closer to Claude Code's native priorities
  13. - reduce routing mistakes around plain agents, teams, tasks, and follow-up workflow
  14. - tighten response behavior so replies stay more direct, restrained, and execution-first
  15. ### Best for
  16. - Claude Code sessions already connected to third-party models
  17. - repositories that expose skills, workflows, MCP resources, or plugins
  18. - users who want fewer detours and less prompt-shape sensitivity
  19. ### Not for
  20. - setting up API keys, providers, or model gateways
  21. - exposing tools that Claude Code itself did not surface
  22. - replacing CCSwitch or another provider-mapping layer
  23. - overriding repository rules such as `AGENTS.md`, `CLAUDE.md`, or direct user instructions
  24. ## What's new in 0.5.11
  25. Compared with the previous public release (`v0.5.9`), `0.5.11` adds two practical fixes and one behavior tightening pass:
  26. | Area | What changed |
  27. |---|---|
  28. | Task completion | `TaskCompleted` / `TaskUpdate(status=completed)` no longer gets hard-blocked just because the description is thin; the hook now warns without breaking completion-state sync |
  29. | New Claude Code compatibility | The old `Task` subagent tool name is now treated as an alias of `Agent`, covering hooks, capability detection, session continuity, and real-session regression checks |
  30. | Output discipline | Explanation and comparison prompts are less likely to trigger team/task-board demos, and the native style now pushes replies toward concise, direct, low-ceremony wording |
  31. ## Features
  32. - **Native-first routing guidance**: keeps the model closer to Claude Code's surfaced capability order instead of falling back to broad keyword-driven choices.
  33. - **Agent and team guardrails**: distinguishes plain workers from real teammate flows, reduces accidental `team_name` pollution, and preserves task-board continuity only when the host proves it exists.
  34. - **Task lifecycle protection**: keeps task creation and task completion checks separate so a completed task is less likely to stay out of sync with the board.
  35. - **Task-to-Agent compatibility**: supports both old and new Claude Code subagent tool names in the same plugin build.
  36. - **Response-style tightening**: reduces over-planning, forced confirmations, tool theater, meta narration, jargon-heavy phrasing, and invitation-style endings.
  37. - **ccstatusline bridge**: backfills status-line usage from transcripts when Claude Code surfaces zeroed usage fields in third-party-model sessions.
  38. ## Quick Start
  39. ### Prerequisites
  40. - Node.js 18 or later
  41. - Claude Code with plugin support
  42. - a working third-party model mapping layer such as CCSwitch, if you are not using native Claude models
  43. ### Install
  44. 1. Clone the repository.
  45. ```bash
  46. git clone https://github.com/hellowind777/hello2cc.git
  47. cd hello2cc
  48. ```
  49. 2. Add the local marketplace entry.
  50. ```bash
  51. claude plugins marketplace add "<repo-path>"
  52. ```
  53. Replace `<repo-path>` with your local `hello2cc` repository path.
  54. 3. Install and enable the plugin.
  55. ```bash
  56. claude plugins install hello2cc@hello2cc-local
  57. claude plugins enable hello2cc@hello2cc-local
  58. ```
  59. 4. Reload Claude Code.
  60. ```bash
  61. /reload-plugins
  62. ```
  63. ### Verify
  64. Run:
  65. ```bash
  66. claude plugins list
  67. ```
  68. Expected result:
  69. - `hello2cc@hello2cc-local` is installed
  70. - the plugin status is enabled
  71. - new sessions no longer rely on a plugin-shipped `settings.json` to force `agent=hello2cc:native`
  72. ## Recommended configuration
  73. ### Minimal configuration
  74. Good when your model mapping is already handled elsewhere and you only want hello2cc to align behavior:
  75. ```json
  76. {
  77. "mirror_session_model": true
  78. }
  79. ```
  80. ### Stable default agent slot
  81. Good when you want most spawned agents to use the same Claude slot:
  82. ```json
  83. {
  84. "mirror_session_model": true,
  85. "default_agent_model": "opus"
  86. }
  87. ```
  88. Use Claude slot values such as `inherit`, `opus`, `sonnet`, or `haiku` here.
  89. If your real target model is mapped through CCSwitch, keep that mapping in CCSwitch instead of writing third-party aliases into hello2cc.
  90. ## How it works
  91. ```mermaid
  92. flowchart LR
  93. A[Third-party model already connected in Claude Code] --> B[Claude Code surfaces tools, agents, skills, or MCP]
  94. B --> C[hello2cc adds host-aware routing and protocol guardrails]
  95. C --> D[Model chooses a more native path]
  96. D --> E[Less drift in tools, teams, tasks, and reply style]
  97. style A fill:#e3f2fd
  98. style B fill:#fff3e0
  99. style C fill:#e8f5e9
  100. style E fill:#4caf50,color:#fff
  101. ```
  102. ### Main behavior layers
  103. | Layer | What it does |
  104. |---|---|
  105. | Host-state guidance | Surfaces tools, agents, workflows, MCP resources, and continuity state so the model chooses within the real session boundary |
  106. | Pre/post tool guardrails | Normalizes inputs, strips placeholders, records failure memory, and keeps deterministic checks fail-closed |
  107. | Native style shell | Keeps replies closer to Claude Code's direct, restrained, execution-first style |
  108. ## Troubleshooting
  109. ### The plugin seems inactive
  110. Check the following:
  111. 1. reload Claude Code or reopen the session
  112. 2. confirm the plugin is installed and enabled
  113. 3. if you updated a local clone, reinstall it cleanly
  114. ### `hello2cc:native` still shows after disable or reload
  115. Claude Code can keep thread-level agent state across resumed sessions.
  116. Current versions no longer ship a plugin-side `settings.json` that force-selects `hello2cc:native`, so a clean reinstall and a fresh session should stop new unintended injection.
  117. ### `TaskCompleted` or `TaskUpdate(status=completed)` still gets stuck
  118. Upgrade to `0.5.11` or later, then reload the plugin.
  119. Current versions downgrade thin completion descriptions from a hard block to a warning-first guard, so task-board state can still move to completed.
  120. ### New Claude Code builds look like `Agent` and `Task` no longer match
  121. Upgrade to `0.5.11` or later.
  122. hello2cc now treats `Task` as an alias of `Agent` across its internal compatibility surface.
  123. ### CCS + sub2api + codex still shows `Team "default" does not exist`
  124. Current versions reduce accidental team injection for explanation, comparison, and capability prompts.
  125. If the problem remains, inspect the Claude Code debug log first and confirm whether the upstream model or proxy is still emitting a real tool input such as `team_name: "default"` or `name + team_name`.
  126. ### Terminal output is still not streaming
  127. This repository does not currently show a plugin-side code path that explicitly disables Claude Code streaming.
  128. If the issue remains, check:
  129. 1. whether `sub2api` buffers streaming responses into a single payload
  130. 2. whether your CCS Anthropic endpoint and Responses endpoint both expose true streaming passthrough
  131. 3. whether `claude --debug-file <path>` already shows the upstream response arriving as non-streaming
  132. ### `ccstatusline` still shows zero usage
  133. Use the bridge command described in [`docs/ccstatusline.md`](./docs/ccstatusline.md).
  134. The bridge reads transcript usage, supports `agentId` / `agent_id` / `agent.id` and direct subagent transcript paths, and backfills `context_window` fields only when Claude Code left them empty or zero.
  135. ## Documentation
  136. - [中文接入说明:ccstatusline 兼容桥接](./docs/ccstatusline.md)
  137. - [Claude Code 重构方案对齐审计](./docs/claude-code-refactor-alignment-audit.md)
  138. - [更新日志](./CHANGELOG.md)
  139. ## FAQ
  140. <details>
  141. <summary><strong>Does hello2cc replace CCSwitch?</strong></summary>
  142. No. CCSwitch or another mapping layer should continue to own provider and model mapping. hello2cc only shapes behavior after the model is already running inside Claude Code.
  143. </details>
  144. <details>
  145. <summary><strong>Does it expose tools that Claude Code did not expose?</strong></summary>
  146. No. It only helps the model choose and use capabilities that are already present in the current session.
  147. </details>
  148. <details>
  149. <summary><strong>Do I need to switch an output style manually?</strong></summary>
  150. Usually not. The plugin ships an output style and a native agent option, but normal installation should not require an extra manual entry point.
  151. </details>
  152. <details>
  153. <summary><strong>Does every multi-agent task become a team?</strong></summary>
  154. No. One-shot parallel work can stay on ordinary agent paths. Real team workflow is reserved for cases that need persistent task-board, ownership, or handoff semantics.
  155. </details>
  156. <details>
  157. <summary><strong>Can hello2cc enforce its wording rules over repository instructions?</strong></summary>
  158. No. User instructions, Claude Code host rules, and repository rules still win. hello2cc only tightens the default behavior inside the remaining space.
  159. </details>
  160. ## Support
  161. - Issues: https://github.com/hellowind777/hello2cc/issues
  162. - Releases: https://github.com/hellowind777/hello2cc/releases
  163. ## License
  164. This repository is licensed under the [Apache-2.0 License](./LICENSE).
  165. See [LICENSE](./LICENSE) for full details.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.