Interactive worksheet
Agent Skills Planning Worksheet
Turn a fuzzy repeatable workflow into a practical agent skill brief, Codex goal prompt, SKILL.md outline, review checklist, and downloadable planning record.
Skill brief builder
Fill in the parts an agent needs before it can work well
The outputs update as you type. Use short plain-language notes first; polish the skill after the workflow is clear.
Generated outputs
Copy or download the working materials
Codex /goal prompt
SKILL.md outline
Review checklist
Quality gate
Review the skill before using it on real work
Scope and trigger
0 / 3
Scope and trigger
Context and tools
0 / 3
Context and tools
Boundaries and quality
0 / 3
Boundaries and quality
Testing and release
0 / 3
Testing and release
Examples and red flags
Use a skill when repetition and judgment matter
Good skill fit
A repeatable workflow with known context sources, safety rules, quality examples, and checks: launch QA, PR review, content publishing, security triage, or workbook generation.
Bad skill fit
A one-time question, vague brainstorming task, or workflow whose rules are still unknown. Start with a normal prompt until the repeatable pattern appears.
Red flags
The skill asks the agent to edit production directly, includes secrets, skips inspection, installs tools without review, or lets AI approve its own risky output.
Strong examples
Show the agent what a good result looks like, what a weak result looks like, and what evidence must be included before the work is considered complete.
Rollout checklist
Ship the skill like a small operating procedure
- Test the skill on a safe sample task before using it on production or client work.
- Compare output against the review checklist and revise unclear instructions.
- Document which agent surfaces, repos, or teams may use it.
- Keep a rollback path: remove the skill, disable activation, or restore the prior version.
- Review the skill after real use and update examples, tests, and boundaries.
Related Academy tools
FAQ
What is an agent skill?
An agent skill is a reusable instruction package that tells an AI agent when to activate, what context to gather, which tools or files to use, what quality bar to meet, and how to verify the result.
When should I create a skill instead of a normal prompt?
Create a skill when the workflow repeats, has quality or safety rules, needs specific files or tools, or benefits from consistent tests and review steps. Use a normal prompt for one-off work.
What makes a skill useful?
A useful skill has a clear trigger, a narrow outcome, required context, constraints, examples, verification steps, and a human approval gate for risky work.
Can a beginner write a skill?
Yes. Start by describing the workflow in plain language, then add the examples, do-not-change rules, tests, and rollback notes that would help another person do the work safely.
Should a skill include secrets or private data?
No. Skills should reference placeholder names, environment variable names, or approved docs. Do not include real keys, tokens, passwords, client records, private URLs, or confidential material.