skill-best-practices
Authors and structures professional-grade agent skills following the agentskills.io spec. Use when creating new skill directories, drafting procedural instructions, or optimizing metadata for discoverability. Don't use for general documentation, non-agentic library code, or README files.
- License 许可
- MIT
- 最近更新
- 2026-07-16
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- skill-best-practices
- metadata
- { "author": "Pedro Nauck", "github": "https://github.com/pedronauck", "repository": "https://github.com/pedronauck/skills" }
- description
- Authors and structures professional-grade agent skills following the agentskills.io spec. Use when creating new skill directories, drafting procedural instructions, or optimizing metadata for discoverability. Don't use for general documentation, non-agentic library code, or README files.
02 资源文件 3
- 资产文件assets/SKILL.template.md974 B
- 参考文档references/checklist.md2.7 KB
- 脚本scripts/validate-metadata.py2.1 KB
03 SKILL.md 正文
Skill Authoring Procedure
Follow these steps to generate a skill that adheres to the agentskills.io specification and progressive disclosure principles.
Bundled Path Rule
Resolve bundled paths relative to the directory that contains this SKILL.md. When invoking a helper from another working directory, expand <skill-best-practices-dir> to that directory first so the command is unambiguous.
Step 1: Initialize and Validate Metadata
- Define a unique
name: 1-64 characters, lowercase, numbers, and single hyphens only. - Draft a
description: Max 1,024 characters, written in the third person, including negative triggers. - Execute Validation Script (read-only helper): Run the validation script to ensure compliance before proceeding:
python3 <skill-best-practices-dir>/scripts/validate-metadata.py --name "[name]" --description "[description]" - If the script returns an error, self-correct the metadata based on the
stderroutput and re-run until successful.
Step 2: Structure the Directory
- Create the root directory using the validated
name. - Initialize the following subdirectories:
scripts/: For tiny CLI tools and deterministic logic.references/: For flat (one-level deep) context like schemas or API docs.assets/: For output templates, JSON schemas, or static files.
- Ensure no human-centric files (README.md, INSTALLATION.md) are created.
Step 3: Draft Core Logic (SKILL.md)
- Use the template in
assets/SKILL.template.mdas the starting point. - Write all instructions in the third-person imperative (e.g., "Extract the text," "Run the build").
- Enforce Progressive Disclosure:
- Keep the main logic under 500 lines.
- If a procedure requires a large schema or complex rule set, move it to
references/. - Command the agent to read the specific file only when needed: "Read references/api-spec.md to identify the correct endpoint."
Step 4: Identify and Bundle Scripts
- Identify "fragile" tasks (regex, complex parsing, or repetitive boilerplate).
- Outline a single-purpose script for the
scripts/directory. - Ensure the script uses standard output (stdout/stderr) to communicate success or failure to the agent.
- In
SKILL.md, reference bundled helpers by a path that is unambiguous from the agent's working directory (the same path the agent will type to invoke them), rather than barescripts/helper.pysnippets that silently depend on CWD. - Label each helper as read-only, bootstrap, or mutating so downstream agents know whether the helper only inspects state or changes it.
Step 5: Final Logic Validation
- Review the
SKILL.mdfor "hallucination gaps" (points where the agent is forced to guess). - Verify all file paths are relative and use forward slashes (
/). - Verify every referenced helper path is unambiguous from the agent's working directory and that the helper's role (
read-only,bootstrap, ormutating) is stated. - Read
references/checklist.mdand cross-reference every checklist item against the final output.
Error Handling
- Metadata Failure: If
scripts/validate-metadata.pyfails, identify the specific error (e.g., "STYLE ERROR") and rewrite the field to remove first/second person pronouns. - Context Bloat: If the draft exceeds 500 lines, extract the largest procedural block and move it to a file in
references/.
04 同仓库更多 Skills
to-prompt-opus
Use when preparing or rewriting prompts for Claude Opus 4.8, Anthropic API agents, XML-structured prompts, long-context document work, coding agents, research synthesis, frontend/design tasks, or model-specific prompt optimization. Do not use for GPT/OpenAI prompts or raw context packaging without Opus-specific tuning.
to-prompt-gpt
Use when preparing or rewriting prompts for GPT-5.5, OpenAI API agents, Responses workflows, coding agents, research synthesis, support assistants, or model-specific prompt optimization. Do not use for Claude/Opus prompts or raw context packaging without GPT-specific tuning.
plan-refactor-qa
Plans and runs qa-report before executing Compozy refactor tasks. Use when the user provides a refactor or issue slug and wants Codex to inspect the matching .compozy/tasks directory, related PRD/TechSpec/ADR/task artifacts, affected code, and existing tests, then invoke $qa-report with a repository-grounded prompt. Do not use to run tests, execute qa-execution, implement code, or perform post-refactor verification.
multi-tool-branch-review
Orchestrates a multi-tool branch review of the current git branch against a base branch. By default it runs two independent thermo-nuclear maintainability audits, one on Codex GPT-5.5 xhigh and one on Claude Opus xhigh, then consolidates the findings into a single remediation plan via writing-plans. If the user explicitly asks for native reviews, it can add the Claude native review pass, the Codex native review pass, or both. Use when the user wants a deep cross-model review of a branch or PR before merge, or asks to review this branch with multiple tools or models. Encodes known infra failure modes and their fixes (missing Claude ACP adapter, Codex MCP auth hangs, false-positive report files, long jobs orphaned inside subagents). Do not use for reviewing a single uncommitted snippet, for non-git work, for running only one reviewer, or when the user only wants a quick lint or test run.
create-pr-with-template
Create a GitHub Pull Request with gh CLI and prefer repository PR templates automatically. Use this whenever the user asks to open/create a PR, especially when template compliance is needed. If multiple templates exist, ask the user which template to use.
ui-craft
Provides guardrails for user-facing UI work: usability heuristics, accessibility floors, design-system discipline, component states, microcopy, motion, dark mode, responsive behavior, and human-AI UX. Use when designing, generating, reviewing, or refactoring visible product surfaces such as components, pages, dashboards, forms, dialogs, loading/empty/error states, or AI interfaces. Do not use for backend-only work, infrastructure, CLI/TUI design, or pure documentation editing.