← 返回 Skills 市场
sheepxux

Somnia

作者 SheepXu · GitHub ↗ · v0.4.3 · MIT-0
cross-platform ⚠ suspicious
34
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install somnia
功能描述
Run overnight skill health reviews, replay-case availability checks, feedback triage, and proposal-only maintenance reports for OpenClaw agents. Use when the...
使用说明 (SKILL.md)

Somnia

Overview

Somnia is the sleep-cycle maintenance layer for OpenClaw skills. It provides a repeatable review workflow that checks installed skills during quiet hours, summarizes health risks, and writes proposal artifacts without silently mutating runtime skills.

Current version: v0.4.3 "Standalone Safety".

Trigger Cues

Use this skill when the user mentions:

  • nightly skill review
  • sleep-time maintenance
  • skill health report
  • skill bug scanning
  • replay regression check
  • feedback-driven upgrade
  • proposal-based update
  • Somnia

Default Workflow

  1. Confirm the review scope: managed skills, feedback-related skills, or all installed skills.
  2. Run lightweight package validation, feedback summary, and replay-case availability checks for each selected skill.
  3. Write JSON and Markdown health reports under the configured learning/report directory.
  4. Write proposal artifacts only when feedback or quality gates justify the change.
  5. Hand proposal artifacts to Skill Forge or a human maintainer before any install decision.
  6. Keep simulated evaluation details hidden from user-facing reports.

Output Contract

The final answer or artifact should include:

  • Review scope and schedule assumption
  • Skills checked and health summary
  • Issues found, grouped by skill
  • Update candidates proposed or blocked
  • Replay and hidden-evaluation pass/fail summary
  • Next action: no-op, review proposal, approve install, or adjust schedule

Quality Gates

  • Never auto-install skill changes; Somnia writes proposals and reports only.
  • Keep hidden evaluation and replay case details out of user-facing Telegram reports.
  • Redact feedback-derived content before it becomes a replay case or report item.
  • Prefer proposal files and manifests over direct mutation of installed skills.
  • Keep Somnia self-contained; do not execute out-of-package Skill Forge code.

Resources

References:

  • references/somnia-architecture.md
  • references/schedule-and-policy.md

Scripts:

  • scripts/nightly_skill_review.py
  • scripts/schedule_nightly_review.py
安全使用建议
Before installing or scheduling Somnia, keep the review scope narrow, avoid enabling feedback-driven proposals until skill-name path validation is fixed, and only use --apply or Telegram reporting after reviewing the generated plist and destination chat. Inspect the full installed scripts if possible, because the provided nightly review source was partially truncated in the review artifact.
功能分析
Type: OpenClaw Skill Name: somnia Version: 0.4.3 The Somnia skill bundle is a maintenance utility designed for scheduled health reviews and reporting of OpenClaw skills. It includes scripts for validating skill metadata, generating health reports, and scheduling recurring tasks on macOS via LaunchAgents (scripts/schedule_nightly_review.py). While the tool can send data to a Telegram bot (scripts/nightly_skill_review.py), this behavior is documented, requires user-provided environment variables, and is restricted to health summaries. The bundle explicitly follows a 'proposal-only' model to prevent unauthorized code execution or auto-updates, showing no signs of malicious intent or hidden exfiltration.
能力评估
ℹ Purpose & Capability
The stated purpose is coherent: it reviews OpenClaw skills, feedback, and replay availability, then writes reports/proposals rather than installing changes.
ℹ Instruction Scope
The workflow asks to confirm scope and says not to auto-install, but supported scopes include feedback-driven and all-skill review, which can cover broad local OpenClaw workspace data.
ℹ Install Mechanism
There is no install spec, but included helper scripts can be run manually; the scheduler is macOS LaunchAgent-specific despite no OS restriction in metadata.
⚠ Credentials
The visible code uses feedback-derived skill identifiers directly in paths, which can escape the intended skill/proposal directories if stored feedback is malformed or poisoned.
ℹ Persistence & Privilege
A scheduled LaunchAgent can be installed only with an explicit --apply flag and KeepAlive is false, but enabling it means the review script runs later without per-run approval.
如何使用
  1. 确保已安装 OpenClaw（本地或 Docker 部署）
  2. 在对话框中输入安装命令：/install somnia
  3. 安装完成后，直接呼叫该 Skill 的名称或使用 /somnia 触发
  4. 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
v0.4.3
Make Somnia standalone and proposal-only: no out-of-package runtime execution, no auto-install modes, and clearer report-only boundaries.
v0.4.2
Initial companion preview for overnight skill health review, replay-aware maintenance reports, and proposal-based Skill Forge updates.
元数据
Slug somnia
版本 0.4.3
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 2
常见问题

Somnia 是什么？

Run overnight skill health reviews, replay-case availability checks, feedback triage, and proposal-only maintenance reports for OpenClaw agents. Use when the... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 34 次。

如何安装 Somnia？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install somnia」即可一键安装，无需额外配置。

Somnia 是免费的吗？

是的，Somnia 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

Somnia 支持哪些平台？

Somnia 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 Somnia？

由 SheepXu（@sheepxux）开发并维护，当前版本 v0.4.3。

推荐 Skills
self-improving agent
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⬇ 463283 · by pskoett
Skill Vetter
Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
⬇ 259410 · by spclaudehome
Polymarket
Query Polymarket prediction markets. Check odds, find trending markets, search events, track price movements.
⬇ 232503 · by joelchance
Self-Improving + Proactive Agent
Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...
⬇ 200423 · by ivangdavila
Github
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
⬇ 191113 · by steipete
ontology
Typed knowledge graph for structured agent memory and composable skills. Use when creating/querying entities (Person, Project, Task, Event, Document), linkin...
⬇ 190156 · by oswalpalash

💬 留言讨论