Use Skill Creator by Anthropic to turn a recurring task into a reusable Skill and improve it through examples. Define when the Skill should activate, draft its instructions, compare test runs, and use feedback to refine the workflow and its description.
Start with a recurring method or an existing Skill that behaves inconsistently. The request can examine real examples, missing instructions, and activation boundaries, producing a reviewable revision and evaluation plan before the guidance is treated as reusable.
Provide sample inputs, the expected output, and examples of good and poor results.
Copy the request into a new Floatboat conversation and provide the required material.
Download FloatboatPreview and copy the request. It is not sent automatically from this page.
Reference: Skill Creator https://github.com/anthropics/skills/blob/main/skills/skill-creator/SKILL.md Read the original documentation and check the current environment and dependencies. Propose any new installation or tool connection before proceeding. My task: Help me turn our weekly customer-feedback summary into a Skill. Define the inputs and output, draft the instructions, and propose test requests before running an evaluation. Expected deliverable: An instruction draft, clear inputs and outputs, and representative evaluation requests.
Skill Creator is Anthropic’s workflow for authoring and evaluating Agent Skills. It starts by clarifying the task, activation conditions, inputs, and expected output, then moves through a draft and representative test prompts. The source includes scripts for inspecting evaluation results and comparing runs, with repeated revision as feedback identifies gaps. This makes it useful for capturing a recurring work process or improving a Skill that activates inconsistently. Its evaluation workflow is written for Claude-based execution and needs the referenced tools to be available. This page describes Anthropic’s implementation; OpenAI and other publishers also use the Skill Creator name for different packages.
A recurring method, an inconsistent Skill, and an inaccurate trigger require different revisions. See how Skill Creator uses real requests and representative runs to improve reusable instructions without confusing activation problems with execution problems.
A recurring task is a useful Skill candidate when its real inputs, manual method, and accepted outputs are already understood. Turn that established process into reusable instructions, then test them with representative user requests. The examples should show whether an agent can reproduce the method beyond the original conversation.
An inconsistent Skill needs successful and failed runs compared against the same acceptance rules. Identify missing steps, unclear expectations, and examples that leave important cases uncovered, then revise the instructions accordingly. Rerun those examples so any improvement can be linked to the failure mode it was meant to address.
A Skill description should match realistic requests while excluding nearby tasks it does not handle. Compare actual user phrasing with positive and negative trigger examples, then narrow ambiguous wording. The resulting activation boundary should help the Skill appear for the intended work without being invoked for a merely related topic.
Two versions of a Skill should be assessed on the same representative requests rather than different demonstrations. Keep the expected behavior and evaluation criteria fixed, then examine where each revision succeeds, fails, or activates unnecessarily. Use the comparison to choose a change with evidence, preserving unresolved cases for another iteration instead of declaring the newer wording automatically better.
Skill creation is an iterative comparison between instructions and real task behavior. Prepare representative requests, draft or revise the guidance, and inspect evaluation results before changing the trigger description or declaring the workflow ready for reuse.
Describe the process, inputs, and expected output. Provide successful examples and any requirements the draft must preserve.
Create the instruction and representative test prompts. Use the source’s supported evaluation tools to inspect how the draft performs.
Compare feedback and test outputs, then change the unclear parts. Rerun relevant cases and check that the description still fits the workflow.
Read another Skill guide and check its original source before using it.
