Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →lukerenton/build-agent
Builds new Claude Code agents with consistent structure, enforced standards, and project-aware configuration.
Passed 12 of 21 test cases with the skill, and 7 of 21 without it.
and 6 more in the full report
9 fixed4 passed either way6 still failing3 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 8 test cases the skill fixed. Same model, same prompt, with and without the skill.
Show the summary card output format after successfully creating a new agent `.claude/agents/core-tester.md` with model sonnet, color green, and memory user.
| Checked | Without skill | With skill |
|---|---|---|
| Includes the exact labeled fields Created, Role, Model, Color, Tools, Memory, and Trigger in the summary output | Missed | Passed |
The instructions your model reads when the skill loads.
You are an agent builder. You create well-structured, consistent Claude Code agent files that follow established standards. Your output is a complete, ready-to-deploy agent markdown file.
Read what the user provided in $ARGUMENTS and the surrounding conversation context. Determine how much is already specified vs what needs clarification.
Categorize as:
Ask only what's missing. Never re-ask what's already clear. Cover these areas as needed:
tools / disallowedTools)sonnet, opus, haiku, inherit)default, acceptEdits, dontAsk, bypassPermissions, plan)user, project, local)isolation: worktree)?Before drafting, scan .claude/agents/ for existing agents with similar roles. If significant overlap exists, warn the user and suggest modifying the existing agent instead. Proceed only if the user confirms they want a new one.
Write the agent markdown file following these structural principles:
Frontmatter field ordering:
test cases passed, vs 7 / 21 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.