← 返回 Skills 市场
643
总下载
1
收藏
5
当前安装
21
版本数
在 OpenClaw 中安装
/install provider-sync
功能描述
Sync provider model lists into OpenClaw config (dry-run preview → confirm → apply). Trigger: /provider_sync
安全使用建议
This skill appears to do what it claims: fetch model lists, normalize them, preview differences, and optionally update your openclaw.json. Before using it: 1) Always run a dry-run first and inspect the generated diff; 2) Use the apply step only in a private session (do not paste API keys in group chats); 3) Note the default config path (/root/.openclaw/openclaw.json) and that the script will create backups and cache under ~/.cache/openclaw/provider-sync — ensure you have backups and appropriate file permissions; 4) Be aware v2 will prune agents.defaults.models by default (use --no-prune-agent-aliases to preserve old behavior); 5) If you rely on provider credentials, prefer providing them via secure local config or CLI in a private context and verify redaction meets your policy (the redaction is best-effort). If you want more assurance, inspect the full script run locally in a sandbox to confirm behavior before letting it modify production configs.
功能分析
Type: OpenClaw Skill
Name: provider-sync
Version: 2.1.5
The provider-sync skill bundle is a utility designed to synchronize AI model lists from upstream providers into the OpenClaw configuration. The core script, scripts/provider_sync.py, demonstrates security-conscious design by implementing a 'dry-run' preview mode, creating automatic backups of configuration files, and enforcing restrictive file permissions (chmod 600). Notably, it includes a redaction function to ensure sensitive fields like API keys and authorization tokens are stripped before being stored in local cache files. The SKILL.md instructions and documentation in the references/ directory align with the stated purpose and emphasize safety boundaries, such as restricting write access in group chat environments.
能力评估
Purpose & Capability
Name/description match the provided script and docs: the tool fetches provider model lists, normalizes fields, shows a dry-run diff, and can apply changes to the local OpenClaw config. No unrelated credentials, binaries, or external installers are requested.
Instruction Scope
SKILL.md limits behavior to fetching provider endpoints, normalizing model metadata, previewing diffs, and writing local openclaw.json (with backup). It explicitly recommends dry-run in group chats and states it redacts secret-like fields before caching. Note: the script can be asked (via CLI params or add flow) to use an apiKey or custom headers — providing such keys is expected for contacting upstream provider APIs but should be done in private and not pasted into public/group chat.
Install Mechanism
No install spec or third-party downloads; this is an instruction+script bundle. The included Python script is executed locally; there is no evidence of external installers or network pulls beyond contacting user-specified provider endpoints.
Credentials
The skill requests no environment variables or platform credentials. It does accept API keys/headers as input when contacting upstream providers (documented as optional). This is proportionate to the purpose, but users should avoid pasting secrets into shared chats and should understand that provider credentials may be used to query upstream endpoints.
Persistence & Privilege
The script writes backups, cache files (~/.cache/openclaw/provider-sync) and by default will read/write the configured OpenClaw config (default path noted as /root/.openclaw/openclaw.json). It also documents a v2 behavior that prunes agents.defaults.models (with a flag to disable pruning). These are expected for a sync tool but are material actions: run dry-run and confirm before 'apply', and be aware it will modify local config and create cache files.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install provider-sync - 安装完成后，直接呼叫该 Skill 的名称或使用
/provider-sync触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v2.1.5
Make auto normalization provider-agnostic: infer generic/gemini/gpt from model family instead of hardcoded provider IDs; scrub environment-specific examples from docs.
v2.1.4
add gpt/codex normalization profile; update docs; fix openai-models example to avoid response-root conflict
v2.1.3
Docs wording cleanup to reduce false-positive suspicious scanning without changing behavior.
v0.1.0
Add fetch progress logs (start/cache-hit/304/fetched/stale-fallback) and label by provider; doc no-cache to avoid stale /models during sync.
v2.1.2
Docs: add restart step (/restart) after apply to complete the workflow; clarify restart is gateway-level and kept as separate explicit step.
v2.1.1
Docs UX: show clickable blue command options (lines start with /provider_sync ...) for provider selection; add /provider_sync add entry.
v2.1.0
UX: text-first docs (no buttons required). Feature: provider-id supports comma list and 'all' (runs multiple providers sequentially, outputs aggregated JSON).
v2.0.4
Docs: clarify no-buttons fallback (provider=cli-usa/cliplus/newapi) and add provider=all one-liner for running all providers without inline buttons.
v2.0.3
Docs: add fallback usage when inline buttons are unavailable: /provider_sync provider=cli-usa.
v2.0.2
Safety: redact secret-like keys before caching upstream /models payload; chmod 600 for backups/updated config where possible. Docs: note v2.0.2 safety fixes.
v2.0.1
Docs: add single-line trigger hint (type /provider_sync) at top of SKILL.md.
v2.0.0
BREAKING: default prune enabled to keep /models counts aligned with provider models. Adds --no-prune-agent-aliases opt-out. Adds cache/timing fetch meta in JSON output; improves Telegram UX guidance.
v1.1.6
Switch license to MIT (standard) + add LICENSE + SPDX metadata
v1.1.5
Clarify local config access, explicit confirmation before writes, provider-scoped updates, config path confirmation, and credential/header expectations. Keep review-first sync flow.
v1.0.1
Revert to 1.0.0 baseline; set latest to stable; deprecate 1.1.x line.
v1.1.4
Add 🚀 emoji to xhigh reasoning hint in dry-run summary.
v1.1.3
Add reasoning-effort hints in dry-run summaries, including xhigh visibility for gpt-5.4; keep hints user-facing without forcing unknown config fields into model objects.
v1.1.2
Only suggest/apply official same-name templates for models missing capability fields; dry-run now separates official-template candidates vs minimal-default fallback.
v1.1.1
Add MIT license, changelog, and README; improve project hygiene and publication completeness.
v1.1.0
Add official model override fallback for missing provider fields; preserve known capabilities during sync; add experimental auto-resolve scaffolding and updated examples.
元数据
常见问题
Provider Sync 是什么？
Sync provider model lists into OpenClaw config (dry-run preview → confirm → apply). Trigger: /provider_sync. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 643 次。
如何安装 Provider Sync？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install provider-sync」即可一键安装，无需额外配置。
Provider Sync 是免费的吗？
是的，Provider Sync 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Provider Sync 支持哪些平台？
Provider Sync 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Provider Sync？
由 Joey-Chen（@c-joey）开发并维护，当前版本 v2.1.5。
推荐 Skills