AGENT SKILL

using-trove

Skill discipline anchor for the Trove marketplace. Loaded once per session; routes the agent to the right trove-* skill before code, planning, review, or git actions.

License 许可
MIT
所属仓库
bravew/trove
最近更新
2026-11-12
Stars
★ 10
Forks
1

01 前言 Front Matter

name
using-trove
metadata
{ "trove-category": "ai-models-agents", "trove-category-label": "AI models & agents" }
description
Skill discipline anchor for the Trove marketplace. Loaded once per session; routes the agent to the right trove-* skill before code, planning, review, or git actions.
when_to_use
session start; skill routing; trove bootstrap
user-invocable
false

02 资源文件 1

03 SKILL.md 正文

<!-- AUTO-GENERATED from SKILL.md.tmpl — do not edit directly --> <!-- Regenerate: bun run build:skills -->

using-trove

User instructions outrank Trove skills, and Trove skills outrank default model behavior. When a user instruction conflicts with a skill, ask a concise clarifying question or name the conflict before continuing. Never silently downgrade a user instruction.

Before code changes, planning work, code review, debugging, skill authoring, delegation, or git actions, check whether a trove-* skill applies. If one applies, name the selected skill once in your response, invoke that skill through the available skill-loading mechanism, and follow it. Do not reinvent the skill guidance inline and do not paraphrase this anchor back to the user.

Routing source: read the generated routing fragment routing.md next to the plugin manifest when this plugin is installed. The repository aggregate lives at docs/routing.md; use it as the fallback when the plugin fragment is not present.

When a workflow skill matches, follow its steps in order; if you skip one, say so in a line rather than dropping it silently. When a trove-principle-* skill shaped a decision, name it and the specific choice it changed (a principle cited with no concrete decision behind it was not applied). Run prose you produce — replies, PR text, release notes — through trove-unslop. Proceed on reversible work and present the result for course-correction; still confirm before irreversible or destructive actions per the hard rules below.

Decision gates use the standard format documented in scripts/lib/decision-gate.ts: context, question, options, and default. Link to or invoke the relevant skill instead of duplicating gate wording from another skill.

Hard rules: do not rewrite tracked files without a user prompt, do not run destructive git without explicit confirmation, do not use --no-verify, and do not claim completion until the relevant verification has actually run or the gap is clearly stated.

04 同仓库更多 Skills