本文へ移動
cccskills
無料GitHub で公開

codex-cc

Delegates coding tasks from Codex to local Claude Code in print mode while preserving Claude's normal runtime customizations by default. Use when: (1) You want Codex to call Claude Code locally, (2) You need Claude-side skills, plugins, MCP servers, custom commands, CLAUDE.md rules, or workspace settings to stay active, (3) You want resumable Claude sessions from Codex via SESSION_ID, (4) You need thin `claude mcp` / `claude plugin` passthrough from the same bridge.

インストール方法を見る

含まれるファイル(5)

  • SKILL.md3.6 KB
  • README.md4.0 KB
  • ruff.toml656 B
  • scripts/claude_bridge.py21.0 KB
  • tests/test_claude_bridge.py24.7 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Quick Start

python scripts/claude_bridge.py --cd "/path/to/project" --PROMPT "Analyze auth flow"

Output: JSON with success, SESSION_ID, agent_messages, stream_file (the raw Claude JSONL stream), and optional all_messages, stderr, or error.

Runtime Contract

  • Default execution uses claude -p and does not force --safe-mode, --bare, --disable-slash-commands, or --strict-mcp-config.
  • Claude Code therefore keeps its normal loading path for trusted-workspace customizations such as CLAUDE.md, skills, plugins, MCP servers, custom commands, and rules.
  • claude -p skips the interactive trust dialog and silently ignores invalid settings files, so use this only in workspaces you already trust and whose .claude settings already validate.
  • The bridge requests stream-json, persists each record immediately, and only treats Claude's final result record as a completed answer. Intermediate assistant records never mask a failed or incomplete turn.
  • On Windows the PROMPT is delivered through stdin when claude resolves to a .cmd/.bat shim, so the prompt never passes through cmd.exe quoting and cannot hit the command-line length limit.

Parameters

usage: claude_bridge.py [-h] --PROMPT PROMPT --cd CD [--SESSION_ID SESSION_ID]
                        [--model MODEL]
                        [--permission-mode {,acceptEdits,auto,bypassPermissions,manual,dontAsk,plan}]
                        [--dangerously-skip-permissions] [--timeout TIMEOUT]
                        [--stream-file STREAM_FILE] [--return-all-messages]
                        {mcp,plugin} ...

options:
  --PROMPT PROMPT                Instruction for the task to send to Claude Code.
  --cd CD                        Workspace root for Claude Code (cwd + --add-dir).
  --SESSION_ID SESSION_ID        Resume a conversation by session UUID.
  --model MODEL                  Claude model override. Aliases resolve against the active endpoint.
  --permission-mode ...          Claude permission mode override.
  --dangerously-skip-permissions Bypass Claude permission checks.
  --timeout TIMEOUT              Bridge-level timeout in seconds. Omit for no bridge deadline.
  --stream-file STREAM_FILE      Raw Claude stream-json JSONL destination.
  --return-all-messages          Include parsed stream records in the result.

subcommands:
  mcp                            Thin passthrough to `claude mcp`.
  plugin                         Thin passthrough to `claude plugin`.

Sessions

Capture SESSION_ID from the first successful response and reuse it for follow-ups. SESSION_ID is empty when a run never established a session (claude produced no stream output): resume only with an id the bridge actually returned.

# New Claude session
python scripts/claude_bridge.py --cd "/project" --PROMPT "Inspect failing tests"

# Resume the same Claude session
python scripts/claude_bridge.py --cd "/project" --SESSION_ID "uuid-from-response" --PROMPT "Now propose the fix"

Passthrough

python scripts/claude_bridge.py mcp list
python scripts/claude_bridge.py plugin list

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

ace-tool

無料

Semantic codebase search, code indexing, and prompt enhancement via standalone CLI. Use when: (1) Semantic code search with natural language queries, (2) Code indexing for remote codebase retrieval, (3) Prompt enhancement with codebase context, (4) Before grep/find/glob operations for better accuracy, (5) Complex requirements clarification, (6) Large codebase navigation. Triggers: "search context", "enhance prompt", "find code that", "index project", "clarify requirements". IMPORTANT: Always use ace-tool BEFORE grep/find/glob for semantic-level code location.

日本語の概要は準備中です。原文の説明を表示しています。

Dianel555/DSkills652026年10月12日 更新

Incremental LLM-friendly wiki generator for Obsidian note vaults. Use when: (1) Building wiki from notes, (2) Ingesting notes to wiki, (3) Obsidian LLM wiki, (4) Incremental knowledge base management. Triggers: 'build wiki from notes', 'ingest notes to wiki', 'Obsidian LLM wiki', 'incremental knowledge base'.

日本語の概要は準備中です。原文の説明を表示しています。

Dianel555/DSkills652026年10月12日 更新

Self-evolution workflow for the agent. Before substantive work, recall past outcomes from evolution memory; while editing, detect improvement signals; at task end, record the outcome; when reusable, distill or search the EvoMap network for proven genes/capsules. Use when the user starts non-trivial work (a feature, a fix, a refactor) or asks the agent to "evolve", "learn from this", or "remember how this went".

日本語の概要は準備中です。原文の説明を表示しています。

Dianel555/DSkills652026年10月12日 更新

cc-agy

無料

Delegates coding/research tasks to the Google Antigravity CLI (`agy`) for external-model execution (Gemini 3.x, Claude Sonnet/Opus 4.6, GPT-OSS). Replaces the broken `collaborating-with-gemini` skill. Use when: (1) External-model delegation via Antigravity, (2) Multi-model prototyping (switch model per call), (3) Backend/logic implementation, (4) Algorithm design and optimization, (5) Bug analysis and debugging, (6) API/database code generation, (7) Code review. Triggers: "delegate to agy", "use Antigravity", "external model", "agy", "Gemini 3.5", "Claude Sonnet 4.6", "GPT-OSS". IMPORTANT: Always request unified diff patches only. Supports multi-turn sessions via SESSION_ID.

日本語の概要は準備中です。原文の説明を表示しています。

Dianel555/DSkills652026年10月12日 更新

cc-codex

無料

Delegates coding tasks to Codex CLI for prototyping, debugging, and code review. Use when: (1) Backend/logic implementation, (2) Algorithm design and optimization, (3) Bug analysis and debugging, (4) API/database code generation, (5) Code quality review and refactoring. Triggers: "implement algorithm", "debug", "analyze code", "backend task", "API implementation", "optimize performance", "refactor", "generate prototype", "code review". IMPORTANT: Default sandbox="read-only" (analysis/review/patch generation); use sandbox="workspace-write" only when codex must run tests/builds to verify. Always request unified diff patches only. Supports multi-turn sessions via SESSION_ID.

日本語の概要は準備中です。原文の説明を表示しています。

Dianel555/DSkills652026年10月12日 更新

context7

無料

Fetch current library and API documentation from Context7 when a coding answer needs authoritative, up-to-date examples or version-specific behavior.

日本語の概要は準備中です。原文の説明を表示しています。

Dianel555/DSkills652026年10月12日 更新

Dianel555 のスキルをすべて見る

このスキルの問題を報告する