Agent Prompt Library

Break the requested outcome into the smallest reliable workflow. For each step define the responsible role, required input, output artifact, decision rule, failure state, and human review point. Identify which steps need current data or external tools. Prefer a simple assisted workflow before proposing full automation.

Evaluate this cross-border product opportunity for a solo operator. Define the buyer and job-to-be-done, collect verifiable demand and competitor signals, estimate unit economics using explicit assumptions, separate facts from hypotheses, and propose three low-cost validation tests. Do not invent sales, traffic, or market-size numbers.

Audit this independent-store product page from the buyer's decision sequence. Evaluate message clarity, offer, proof, objections, delivery, returns, mobile friction, and analytics coverage. Cite visible evidence, distinguish defects from hypotheses, and return no more than seven experiments with metric, effort, and risk.

Build an SEO article outline for the target query. Infer the dominant intent from observable search results, define the reader's decision, assign one job to every section, list evidence or examples required, identify claims that need current sources, and propose internal links. Avoid generic headings and keyword repetition.

Analyze the workflow I provide. Define its narrow job, triggers, required inputs, output contract, safety boundaries, and pass/fail checks. Produce a concise SKILL.md and move optional detail into referenced resources. Do not invent tools or automation that the workflow has not validated.

Review the supplied Agent Skill as an installer and evaluator. Check trigger precision, scope, required context, tool permissions, destructive actions, external data handling, output contract, failure behavior, and test coverage. Separate confirmed issues from suggestions. Return blocking issues first, then a minimal patch plan.