← 返回 Skills 市场
730
总下载
0
收藏
2
当前安装
7
版本数
在 OpenClaw 中安装
/install preqstation
功能描述
Delegate PREQSTATION coding tasks to Claude Code, Codex CLI, or Gemini CLI with PTY-safe execution (workdir + background + monitoring). Use when building, re...
安全使用建议
This skill is largely coherent with its stated purpose: it runs local coding CLIs in per-task git worktrees and keeps a MEMORY.md mapping of projects. Before installing, verify the following: (1) you have and trust the local CLI binaries (claude/codex/gemini) and git — the skill expects them though the registry metadata does not list git or the env var; (2) understand and approve that the skill will read and update MEMORY.md in the repository (it will store absolute paths you provide); (3) ensure mappings do not point to sensitive system directories and that OPENCLAW_WORKTREE_ROOT (default /tmp/openclaw-worktrees) is acceptable; (4) consider disabling autonomous invocation if you do not want the agent to autonomously launch local CLIs or modify files; (5) ask the publisher to fix the package metadata (declare git as a required binary and OPENCLAW_WORKTREE_ROOT in requires.env) so the manifest matches runtime expectations. If any of these are unacceptable or the metadata remains inconsistent, treat the skill as untrusted.
功能分析
Type: OpenClaw Skill
Name: preqstation
Version: 0.1.8
The skill automates coding tasks using external CLIs (Claude, Codex, Gemini) and explicitly instructs the agent to use high-risk flags such as '--dangerously-skip-permissions' and '--dangerously-bypass-approvals-and-sandbox' to bypass interactive security prompts during background execution. While SKILL.md implements safety mitigations like mandatory git worktrees and directory blacklisting (e.g., ~/clawd/), the intentional bypass of underlying tool sandboxes and the use of background PTY sessions for arbitrary code modification represent significant high-risk capabilities.
能力评估
Purpose & Capability
The name/description (delegate local coding tasks to claude/codex/gemini CLIs with worktrees/background/monitoring) aligns with the SKILL.md instructions. However the published registry metadata and the SKILL.md disagree: the registry lists no required binaries and no env vars, but SKILL.md requires at least one engine binary (claude/codex/gemini) and also requires git on PATH. This metadata mismatch is unexpected and worth correcting before trusting the package manifest.
Instruction Scope
The instructions instruct the agent to read and update MEMORY.md (project-to-path mappings), create git worktrees, and launch local CLIs in those worktrees using pty/background/monitoring. These actions are coherent for the purpose, but they involve modifying repository files and running arbitrary local binaries (claude/codex/gemini and any commands run via them) — a user should explicitly consent to the skill modifying MEMORY.md and executing in local workspaces. SKILL.md also references OPENCLAW_WORKTREE_ROOT (an environment variable) and enforces absolute-path validation; the skill both reads and writes local filesystem state, which increases risk if mappings point to sensitive directories.
Install Mechanism
Instruction-only skill with no install spec and no code files — lowest install risk. Nothing is downloaded or executed from remote URLs by the skill package itself.
Credentials
The skill does not request cloud credentials or secrets, which is appropriate. However SKILL.md expects an optional OPENCLAW_WORKTREE_ROOT env var and requires git and engine binaries, while the registry metadata declares none. This inconsistency between declared requirements and runtime instructions is a proportionality/manifest problem: the skill will fail or behave unexpectedly if those runtime prerequisites are not present, and the package manifest does not advertise them.
Persistence & Privilege
The skill is not always-enabled and uses normal autonomous invocation. It writes to its own MEMORY.md mapping file in the repository and creates per-task git worktrees under a configurable root; it does not claim system-wide or other-skills privileges. Still, autonomous invocation combined with the ability to launch local binaries and modify workspace mappings increases the blast radius — consider restricting autonomous invocation if you are concerned.
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 preqstation - 安装完成后，直接呼叫该 Skill 的名称或使用
/preqstation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
v0.1.8
sync skill updates
v0.1.7
preqstation 0.1.7
- Added runtime prerequisite documentation for required binaries and environment variables.
- Introduced explicit preflight safety checks before code execution (e.g., verifying git and the requested engine are available).
- Clarified execution safety gates: enforce worktree use, forbid running in primary or protected paths, and improve handling of permission/sandbox bypass flags.
- Expanded documentation for why "dangerously-*" flags are retained and emphasized failure handling if the environment blocks these flags.
- No functional code behavior changes; documentation and process clarifications only.
v0.1.5
preqstation 0.1.5
- Added README.md with usage and skill details.
- No code or logic changes; documentation added only.
v0.1.4
preqstation 0.1.4
- Added MEMORY.md: now uses a project lookup table for path and key resolution.
- Project path resolution tightened: exact (not partial/fuzzy) project key match required in MEMORY.md.
- Skill will always prompt user to confirm or create mappings for new/unmapped project keys before executing tasks.
- Streamlined trigger logic: now highest priority if user mentions `/skill preqstation`, `preqstation`, or `preq`.
- Worktree and branch naming standardized: only uses `codex/<project_key>` (no ticket/task suffix).
- Documentation clarified on mapping updates, execution flow, and stricter user confirmation before any workspace is used.
v0.1.3
**Major update: Adds strict PTY execution, worktree isolation, and safer project workspace handling for coding agents.**
- Enforces PTY-safe execution (`pty:true`) for all coding agent launches.
- Requires all code changes to run in a git worktree (never the primary checkout).
- Introduces explicit routing and branch naming conventions tied to project key/task—prevents accidental modifications to sensitive directories.
- Adds interactive flow for missing project mappings, with instant `MEMORY.md` updates on user confirmation.
- Removed one-liner edit and read-only inspection support; this skill is now strictly for orchestrating agent-driven coding tasks.
- Clarifies and strengthens rules for agent orchestration, session monitoring, worktree management, and output policy.
v0.1.2
preqstation v0.1.2 changes:
- Major redesign: Skill now accepts natural-language requests and no longer requires the `preqstation:` prefix.
- Adds support for resolving project paths using a new `MEMORY.md` file with a Projects table.
- Now allows users to add or update project mappings directly by editing `MEMORY.md`.
- Improved extraction of `engine`, `task`, and `cwd` from user input with fallback logic.
- Updates prompt/output templates and execution rules for greater flexibility.
v0.1.1
preqstation 0.1.1
- Adds execution of Claude Code, Codex CLI, or Gemini CLI via OpenClaw for commands starting with "preqstation:".
- Strictly requires `engine`, `cwd`, and `prompt` flags in the input; optionally supports `task`.
- Renders user prompt into a fixed template before execution for consistent context.
- Returns only short completion summaries, masking raw command outputs.
- Triggers exclusively if input starts with "preqstation:".
元数据
常见问题
Preqstation 是什么？
Delegate PREQSTATION coding tasks to Claude Code, Codex CLI, or Gemini CLI with PTY-safe execution (workdir + background + monitoring). Use when building, re... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 730 次。
如何安装 Preqstation？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install preqstation」即可一键安装，无需额外配置。
Preqstation 是免费的吗？
是的，Preqstation 完全免费（开源免费），可自由下载、安装和使用。
Preqstation 支持哪些平台？
Preqstation 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 Preqstation？
由 sonim1（@sonim1）开发并维护，当前版本 v0.1.8。
推荐 Skills