com.codexguild由 codexguild.com 發布官方發布者v1.17.0更新於 Oct 29, 2026
What changed in your stack since the model's cutoff, error fixes, and security-scanned skills/MCPs
CodexGuild 的託管 MCP 服務（api.codexguild.com），由 codexguild.com 發布；需要 API 金鑰；Streamable HTTP；7 個工具。
暫無 AI 概覽：以下資訊來自註冊中心條目和服務本身。
Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。
閱讀 https://sourceweft.com/mcp/com.codexguild%2Fcodexguild/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。把它新增到你客戶端的 mcpServers 設定中。
{
"mcpServers": {
"codexguild": {
"type": "http",
"url": "https://api.codexguild.com/v1/mcp",
"headers": {
"Authorization": "Bearer <YOUR_TOKEN>"
}
}
}
}Security-audit the skills installed on THIS machine (~/.claude/skills, Claude plugins, ~/.codex/skills, ~/.agents/skills, ~/.hermes/skills, project skills…). Detects pipe-to-shell, credential exfiltration, hidden unicode, prompt injection, agent-config tampering and safety bypasses. Runs locally — nothing is uploaded.
pathsverboseLint the agent instruction files in this project (AGENTS.md, CLAUDE.md, GEMINI.md, .cursorrules…): size per session, missing build/test commands, README copy, generic advice, secrets, machine-specific paths, prompt-injection and other scanner findings. Runs locally — file contents are not uploaded.
filesSecurity-review the GitHub Actions workflows in this project that run coding agents (claude-code-action, run-gemini-cli, codex, opencode, aider…): triggers outside contributors control (pull_request_target, issue_comment, issues) combined with secrets, issue/PR text interpolated into prompts or shells (prompt injection), unrestricted agent tools, broad permissions, unpinned agent actions. Runs locally — nothing is uploaded.
filesFind deprecated and removed APIs in this project's code for the exact versions it runs (from the lockfiles): e.g. `app.mount` deprecated in hono 4.13, with the replacement and where it is used (file:line). Only package names and versions are sent to CodexGuild; the code is searched locally and never uploaded. codexguild_sync runs this automatically.
cwdPattern-scan this project's OWN code for common vulnerability classes (OWASP/CWE): hardcoded secrets, SQL injection via string concatenation, eval/shell injection, insecure deserialization, weak crypto, SSRF, path traversal, and IDOR-prone lookups — with file:line and a fix for each. Pattern matching, not proof: read a finding before changing code. Runs locally — nothing is uploaded. Run this before finishing a feature/PR that touches routes, queries, auth, serialization or file/network I/O; codexguild_sync also runs it once at session start.
cwdScore how well THIS repository is set up for a coding agent (0–100): instruction file quality and whether it names the build/test commands, test and type-check commands, lockfile, secrets kept out of git, CI, safe agent workflows, CodexGuild hooks — with the fixes worth doing, biggest first. Runs locally — nothing is uploaded.
cwdReview this project's recent agent sessions (Claude Code / OpenClaude transcripts on this machine) for where time was lost: the same error hit repeatedly, retry loops, refused tool calls (with allowlist suggestions), edits that missed, files re-read many times, compactions, sessions that ended with unchecked code, MCP servers that never get used — and what to change. Use it when the user asks how to make the agent work better here. Runs locally — transcripts are not uploaded.
allcwddays已滿足適用檢查項的 75%
該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。
請宣告包含 README 的 GitHub 儲存庫，或在 OAuth 受保護資源後設資料中提供 resource_documentation。
請在儲存庫或註冊中心條目中宣告授權條款。