create-agent
Create and update Harness AI agent instances - standalone templates used as building blocks in pipelines. Agents contain a single agent step with connector-driven architecture requiring llmConnector (LLM access) and optional mcpConnectors (GitHub, Slack, Harness platform). Supports runtime inputs and task/rules-based instruction. Use when asked to create an agent, update agent spec, modify agent configuration, automate tasks, perform agentic workflows, build autonomous systems, or work with AI agents. Trigger phrases: create agent, update agent, modify agent spec, AI agent, autonomous agent, agentic pipeline, automation task, automate workflow, Harness agent, code coverage agent, review agent, agentic task.
82
73%
Does it follow best practices?
Impact
96%
3.42xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
Optimize this skill with Tessl
npx tessl skill review --optimize ./plugins/claude/skills/create-agent/SKILL.mdQuality
Discovery
92%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 is a strong skill description that provides detailed technical specifics about Harness AI agent creation, includes explicit 'Use when' guidance, and offers comprehensive trigger phrases. Its main weakness is that some trigger terms ('automate tasks', 'autonomous systems', 'agentic workflows') are generic enough to potentially conflict with other automation-related skills, though the Harness-specific terminology helps mitigate this.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | The description lists multiple specific concrete actions and architectural details: creating/updating agent instances, connector-driven architecture with llmConnector and mcpConnectors (GitHub, Slack, Harness platform), runtime inputs, and task/rules-based instruction. It goes beyond naming a domain and provides detailed implementation specifics. | 3 / 3 |
Completeness | Clearly answers both 'what' (create and update Harness AI agent instances with connector-driven architecture) and 'when' (explicit 'Use when...' clause listing specific scenarios plus a dedicated 'Trigger phrases' section). Both dimensions are thoroughly addressed. | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural trigger terms including 'create agent', 'update agent', 'AI agent', 'autonomous agent', 'agentic pipeline', 'automate workflow', 'Harness agent', 'code coverage agent', 'review agent', and 'agentic task'. These cover a wide range of natural phrases a user would say when needing this skill. | 3 / 3 |
Distinctiveness Conflict Risk | While 'Harness AI agent' is fairly specific, terms like 'automate tasks', 'autonomous systems', 'agentic workflows', and 'AI agent' are quite broad and could overlap with other automation or AI-related skills. The Harness-specific terms help, but the generic automation triggers introduce some conflict risk. | 2 / 3 |
Total | 11 / 12 Passed |
Implementation
55%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill is highly actionable with concrete YAML examples, specific API calls, and a well-structured 5-phase workflow with validation checkpoints. However, it is excessively verbose — repeating configuration patterns multiple times, over-explaining requirements gathering steps that Claude can handle implicitly, and inlining all content into a single monolithic file. The lack of progressive disclosure means the entire ~250+ line document must be loaded every time, wasting significant context window budget.
Suggestions
Reduce verbosity by 40-50%: remove the detailed requirements gathering prompts in Phase 2 (Claude knows how to ask clarifying questions), eliminate redundant YAML fragments that appear in both the 'Default Configuration' section and Phase 3/Example, and cut obvious advice like 'Be specific — avoid vague goals'.
Split into multiple files: move the full Code Review Agent example to EXAMPLES.md, troubleshooting to TROUBLESHOOTING.md, and the detailed default configuration/input types reference to REFERENCE.md, keeping SKILL.md as a concise overview with links.
Consolidate the default configuration section and Phase 3 into a single canonical YAML template — currently the same information (container image, env vars, mcp_servers, inputs) appears three times (defaults, Phase 3 instructions, and full example).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose at ~250+ lines. Phase 2's requirements gathering section extensively explains how to ask clarifying questions and lists obvious examples Claude doesn't need spelled out (e.g., 'Be specific — avoid vague goals like improve code quality'). The default configuration section repeats YAML fragments that are then repeated again in Phase 3 and the full example. Significant redundancy throughout. | 1 / 3 |
Actionability | The skill provides concrete, executable YAML specs, specific MCP tool calls with exact parameters, complete example agent configurations, and precise field names/types. The create/update API calls are copy-paste ready with all required parameters clearly specified. | 3 / 3 |
Workflow Clarity | The 5-phase workflow is clearly sequenced with explicit checkpoints: check existing agents first, gather requirements interactively, generate spec, present for review and wait for confirmation, then create/update. Validation steps are present (check existing, review before creating, verify connectors). The troubleshooting section provides error recovery guidance. | 3 / 3 |
Progressive Disclosure | This is a monolithic wall of text with no bundle files or external references. The default configuration section, full example, best practices, troubleshooting, and critical guidelines could all be split into separate referenced files. Everything is inlined into a single massive document, making it expensive to load for every interaction. | 1 / 3 |
Total | 8 / 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
- harness/harness-ai
- Commit
a000afa
- 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.