Trinity Evolution
/install trinity-evolution
Trinity Evolution
Use this skill to operate and audit the OpenClaw self-evolution loop.
Core Workflow
- Check the local system status command before making readiness claims.
- Run the operational preflight before publishing or reporting a release baseline.
- Confirm there are no pending holdouts, repair failures, or unverified current repairs.
- Treat external validation and explicit validation as promotion gates.
- Reject self-referential proof, internal metrics alone, auto-scored results alone, social-only signals, and stale repair status.
- Summarize user-facing impact in plain language: what OpenClaw improved, why it matters, and how the user should use it.
Evidence Rules
Only claim a capability improvement when the current repair has:
- At least two independent external PASS results.
- Zero external FAIL results for the current repair.
- At least one explicit positive validation.
- Zero pending current-repair holdouts.
- A source-backed or artifact-backed rationale.
If any required evidence is missing, state that the evidence is insufficient and do not claim improvement.
v16.0 Baseline
v16.0 means the loop can:
- Discover candidate OpenClaw capabilities from source-backed direction radar.
- Materialize a candidate into repair and holdout records.
- Ask OpenClaw to answer holdouts.
- Judge outputs with an external or configured OpenClaw judge.
- Promote only after repair-level gates pass.
- Preserve failed repairs and create conservative replacement repairs without fabricating success.
- Produce a user-facing daily progress report.
User-Facing Reporting
Reports should describe OpenClaw's practical capability changes, not internal Trinity mechanics.
Use plain language:
- What OpenClaw can now do better.
- Why that matters to the user.
- How the user should use the new behavior.
- What evidence supports the claim.
Safety And Publishing Notes
Do not include sensitive local data, private operational records, messaging service configuration, machine-specific paths, or generated private state in public artifacts.
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install trinity-evolution - 安装完成后,直接呼叫该 Skill 的名称或使用
/trinity-evolution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
Trinity Evolution 是什么?
Audit and operate the Trinity/OpenClaw evolution loop by validating status, preflight, capability gates, repair fallback, and user progress reports. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 49 次。
如何安装 Trinity Evolution?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install trinity-evolution」即可一键安装,无需额外配置。
Trinity Evolution 是免费的吗?
是的,Trinity Evolution 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Trinity Evolution 支持哪些平台?
Trinity Evolution 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 Trinity Evolution?
由 hollis9087(@hollis9087)开发并维护,当前版本 v16.0.0。