agent-development
This skill should be used when the user asks to "create an agent", "write an agent", "build an agent", or wants to add new agent capabilities to Claude Code.
53
29%
Does it follow best practices?
Impact
96%
1.84xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
Optimize this skill with Tessl
npx tessl skill review --optimize ./plugins/toolkit/skills/agent-development/SKILL.mdQuality
Discovery
37%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This description is heavily skewed toward trigger terms and 'when to use' guidance while completely neglecting what the skill actually does. It reads more like a trigger rule than a skill description. Without any concrete actions or capabilities listed, Claude cannot make an informed decision about whether this skill is appropriate beyond keyword matching.
Suggestions
Add concrete capability descriptions explaining what the skill produces (e.g., 'Scaffolds agent architectures with tool definitions, agent loops, and error handling for Claude Code')
Include a 'what it does' section before the 'Use when...' clause that lists specific actions like generating agent boilerplate, configuring tool schemas, or setting up multi-step workflows
Rewrite in third person voice (e.g., 'Generates agent scaffolding and configurations for Claude Code. Use when...')
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | The description does not describe any concrete actions or capabilities. It only mentions 'create an agent', 'write an agent', 'build an agent', or 'add new agent capabilities' without explaining what the skill actually does (e.g., scaffolds agent code, configures tools, defines agent loops). | 1 / 3 |
Completeness | The description answers 'when' (when the user asks to create/write/build an agent) but completely fails to answer 'what' — there is no explanation of what the skill actually does or produces. The 'what' is very weak, essentially absent. | 1 / 3 |
Trigger Term Quality | The description includes natural trigger phrases users would actually say: 'create an agent', 'write an agent', 'build an agent', and 'add new agent capabilities to Claude Code'. These are realistic user phrasings with good variation coverage. | 3 / 3 |
Distinctiveness Conflict Risk | The term 'agent' provides some specificity, and the mention of 'Claude Code' narrows the domain. However, 'agent capabilities' is broad enough that it could overlap with skills related to tool use, automation, or general Claude Code development. | 2 / 3 |
Total | 7 / 12 Passed |
Implementation
22%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill is fundamentally incomplete for its stated purpose of helping create agents. It describes cosmetic properties (colors) and a behavioral note (permission resilience) but entirely omits the core workflow: how to actually create an agent file, where to place it, what structure it needs, and how to verify it works. The skill reads more like a reference appendix than an actionable creation guide.
Suggestions
Add a complete, executable example of creating an agent from scratch — including full file content, directory path (e.g., `.claude/agents/explore.md`), and the expected file structure.
Include a step-by-step workflow: 1) Create file in correct directory, 2) Write YAML frontmatter + body content, 3) Verify agent appears/works. This is essential for a 'create an agent' skill.
Add a concrete example of agent body content (the markdown instructions after frontmatter) so users understand what makes an effective agent, not just what color to assign it.
Remove or significantly trim the 'Permission Denial Resilience' section — this describes built-in platform behavior rather than actionable creation guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The color conventions table is useful reference material, but the 'Permission Denial Resilience' section explains built-in behavior Claude likely already knows. The 'Important' section about hot-reloading is a good concise note. Overall mostly efficient but has some unnecessary explanation. | 2 / 3 |
Actionability | The skill lacks concrete, executable guidance on how to actually create an agent. It shows a YAML frontmatter snippet for color but doesn't provide a complete agent file example, directory structure, or step-by-step instructions for building an agent. For a skill triggered by 'create an agent,' this is critically incomplete. | 1 / 3 |
Workflow Clarity | There is no workflow or sequence of steps for creating an agent. The skill describes properties of agents (colors, behavior) but never walks through the process of actually building one — no file creation steps, no directory placement, no validation. | 1 / 3 |
Progressive Disclosure | The content is reasonably organized with clear sections and a useful table, but it doesn't reference any external files for deeper topics (e.g., advanced agent patterns, examples). For a skill about creating agents, there's no pointer to example agents or templates. | 2 / 3 |
Total | 6 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Repository
- dwmkerr/claude-toolkit
- Commit
ccd0360
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.