AGENT SKILL

add-skill

Use when building a skill, an agent or a command inside a plugin that exists: which file it is, where it goes, the size cap, the description that makes it fire, and the eval that must pass before it is done. Triggers: add a skill, new skill, create a skill, add an agent, new subagent, add a slash command, pridej skill, pridej agenta, udelej novy skill. A new plugin is add-plugin, a hook is add-hook, a CLI verb is add-tool; whether a skill is any good is skill-standards.

License 许可
GPL-3.0
最近更新
2026-11-13
Stars
★ 0
Forks
0

01 前言 Front Matter

name
add-skill
description
Use when building a skill, an agent or a command inside a plugin that exists: which file it is, where it goes, the size cap, the description that makes it fire, and the eval that must pass before it is done. Triggers: add a skill, new skill, create a skill, add an agent, new subagent, add a slash command, pridej skill, pridej agenta, udelej novy skill. A new plugin is add-plugin, a hook is add-hook, a CLI verb is add-tool; whether a skill is any good is skill-standards.

03 SKILL.md 正文

Add Skill

A skill, an agent and a command are the three prompt files a plugin ships. This is the build order and the hard rules for adding one, because a skill added by feel loads never, loads on everything, or duplicates a neighbour, and none of that shows until it is used. How to write the prose is skill-standards; a whole new plugin is add-plugin.

Steps

  1. Search first: qsm-memory search <concept>, then grep skill descriptions and agent files. A skill that covers it gets edited, and you stop here.
  2. Pick the kind. A skill is knowledge or a procedure the main agent loads on a trigger. An agent is one bounded job in its own context that returns a briefing. A command is a prompt the person types by name.
  3. Pick the plugin that owns the subject. Paths: plugins/<plugin>/skills/<name>/SKILL.md, plugins/<plugin>/agents/<name>.md, plugins/<plugin>/commands/<name>.md.
  4. Write the evals, run them once, and read what the model does unaided: that is the baseline.
  5. Write the description from the eval's prompt, then the body to the shape skill-standards gives.
  6. Run Done means. When a fire case fails, fix the description first: the harness matches it.

Hard rules

  1. The file lives under the plugin that owns the subject, never in root lib/ or another plugin, because a plugin owns its content. Gate: validate-no-cross-plugin-import, review.
  2. Frontmatter carries name and description (skill), plus model (agent), or description (command), because a malformed file is dropped silently. Gate: validate-plugin-contracts.
  3. An agent names its cheapest fitting model, because an unnamed one inherits the session's most expensive. Gate: validate-plugin-contracts.
  4. A SKILL.md or agent stays under 12 KB, aims for 8, with no references/; over the cap is two skills, because a second subject loads where it is not needed. Gate: validate-file-size.
  5. The description opens with the key case, carries Czech and English triggers, stays under 1,024 characters, and names the neighbour that owns the rest, because two matching descriptions are a coin flip. Gate: the fire eval, review for the neighbour.
  6. A skill or agent ships, before the body, a fire case, a behaviour case (scenario) and one negative case per plugin in plugins/<plugin>/evals/, because an unproven skill may never load or may change nothing. The case format, regex samples included, is skill-standards' Evals section. Gate: review.
  7. A skill names a CLI verb on PATH, never a path into this repository or another plugin's files, because a binary runs with no checkout in the room (add-tool owns the rule). Gate: review.

Done means

Every command exits 0, as mise exec -- ...:

  • bun run validate
  • bun run lint
  • bun run typecheck
  • bun test ./ ./.gates

A skill's behaviour is defined through its evals.

Then hand the diff to the strict-reviewer agent of qsm-review with the hard rules as the checklist, and add-plugin's four reasons for every new file. Done means no finding is left.

Red flags

  • When the description is written after the body, rewrite it from the eval's prompt, because the harness matches what users type.
  • When an agent is one nobody delegates to, make it a skill, because an agent costs a context and a model name.

Say it in the reply

One line naming the kind, the plugin and the path (or the existing owner). Then each Done means command with pass or fail, the eval cases with the with and without result, and the reviewer's verdict.

04 同仓库更多 Skills

qsmark-labs
qsmark-labs/qsm-marketplace

worktrunk

Use when working with git worktrees or wt (worktrunk): creating, switching, merging or removing a branch workspace, configuring wt (project or personal config, workflow hooks around switch, commit and merge, commit-message generation, hook approvals), or a wt command that failed. Triggers: make a worktree for this branch, wt merge, wt remove, configure wt, .config/wt.toml, run tests before merge, install deps when a worktree is created, wt hook, wt switch does not cd, nastav wt, wt nefunguje.

★ 0
qsmark-labs
qsmark-labs/qsm-marketplace

web-ladder

Use when a task needs something the internet holds: look something up, search the web, find which page answers a question, read a page, check a library's current docs, a page that needs a browser or shows a captcha. Runs one research loop: searxng search, keywords, cases, search again up to three times, Context7 for a library, Tavily when the keywords will not come, one page in a browser, the person on a wall. Triggers: look this up, search the web, find out online, read this page, check the docs, the page is blocked, captcha, najdi na webu, otevri tu stranku. A multi-round question goes to the qsm-web:web-research agent; one tool's commands are its own skill (searxng, browserbox, opensrc).

★ 0
qsmark-labs
qsmark-labs/qsm-marketplace

searxng

Use when running a web search through the person's own SearXNG instance with qsm-web searxng: choosing a search profile or engines, the instance, time range or language, reading its JSON, a suspended engine. Triggers: search with searxng, qsm-web searxng, search profile, niche search, academic search, IT search, vyhledavaci profil, searxng is down, which engines. The research loop around it is the web-ladder skill.

★ 0
qsmark-labs
qsmark-labs/qsm-marketplace

opensrc

Use when the source code of an open-source library or repository is needed on disk to read: the docs are silent, the behaviour must be checked in code, the version the project pins. Runs qsm-web opensrc get with a package spec or a repository address. Triggers: get the source of this package, clone this library, read the library's code, qsm-web opensrc, stahni zdrojak knihovny. Library docs are the Context7 skill; the research loop is the web-ladder skill.

★ 0
qsmark-labs
qsmark-labs/qsm-marketplace

browserbox

Use when one page needs a real browser through qsm-web browserbox: content built by script, a page that blocks plain fetching, a captcha to solve or hand to the person, a scripted sequence of clicks, a login the person is present for. Triggers: open it in a browser, the page is empty without JavaScript, captcha, take over the browser, run these steps on the page, log in to this site, browserbox. When to reach for a browser at all is the web-ladder skill.

★ 0
qsmark-labs
qsmark-labs/qsm-marketplace

usage

Use when the user wants a review of how they use Claude Code, or setup hands over after a first install: review my Claude Code usage, audit my setup, what does my setup cost, where did my tokens go, clean up my Claude Code, usage report, zkontroluj moje pouzivani. One walkthrough of four stages (machine, work, habits, extensions), questions after each, then one plan. A question about one topic (which hooks are slow) goes to that topic's skill; this session's tokens alone is `/qsm-usage`.

★ 0