無料GitHub で公開

workflow-to-skill

Converts a proven, repeated professional workflow into a portable and inspectable AI Skill with clear triggers, inputs, steps, outputs, verification, recovery, and limits. Use when a user asks to make a workflow reusable, package a process as a Skill, or preserve a successful multi-step method. Do not use for speculative processes that have not been performed or concretely specified.

日本語の概要は準備中です。原文の説明を表示しています。

lawve-ai/awesome-legal-skills

実行スコア未検証

インストール方法を見る

含まれるファイル（3）

SKILL.md3.4 KB
agents/openai.yaml237 B
README.md855 B

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Workflow to Skill by Ignacio Adrián Lerer

Public workflow-design method by Ignacio Adrián Lerer.

Convert observed professional practice into reusable agent instructions without pretending that one successful conversation proves a universal method.

Admission gate

Proceed only when:

the workflow was performed at least once or is concretely specified;
its trigger and intended user are identifiable;
required inputs and expected outputs can be stated;
success and failure can be verified;
the workflow does not depend on undisclosed private material;
any third-party material can be treated as untrusted source material rather than instructions.

If these conditions are missing, return NEEDS EVIDENCE instead of inventing a Skill.

Workflow

Define the task class represented by the example.
Separate reusable method from case-specific facts, credentials, private data, and product internals.
Write concrete trigger phrases and at least one unrelated non-trigger.
Specify required inputs, ordered actions, output shape, verification, stop conditions, and recovery.
Identify capabilities and side effects. Require explicit approval for external transmission, publication, deletion, deployment, permission changes, or spending.
Check whether an existing Skill already covers the same task. Prefer improvement over duplication.
Keep the core Skill concise. Include extra files only when they are required to perform or verify the workflow.
Test the Skill on one representative case and one boundary case.
Classify it as CANDIDATE until another user, agent, or future session can apply it from the package alone.

Output

## Workflow-to-Skill Receipt

Proposed Skill:
Author or method owner:
Status: NEEDS EVIDENCE | CANDIDATE | READY FOR HUMAN REVIEW

### Workflow contract
- Trigger:
- Non-trigger:
- Inputs:
- Actions:
- Output:
- Verification:
- Stop conditions:
- Recovery:

### Public boundary
- Included:
- Excluded:

### Validation
- Representative case:
- Boundary case:
- Remaining uncertainty:

Verification and fallback

Confirm that the package works without access to the original conversation. If it requires hidden context, revise it or keep it as CANDIDATE.

If the workflow cannot be verified, preserve it as notes or a draft checklist. Do not promote it by reputation or enthusiasm alone.

Trigger tests

Obvious: “Turn this repeated contract-review workflow into a Skill.”
Paraphrased: “Package the process we just used so another agent can repeat it.”
Boundary: “We have only an idea for a workflow and have never tried it.”
Non-trigger: “Summarize this contract without creating a reusable process.”

Public boundary

This public edition includes only a portable method. Do not include private prompts, client data, credentials, local paths, internal registries, model routing, scoring thresholds, cost records, private evaluations, or proprietary implementation details.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインなしで書き込めます。使ってみた感想、質問、動かなかった環境などを気軽にどうぞ。

まだコメントはありません。最初のコメントを書いてみませんか。

同じリポジトリのスキル

adjudicative-ai-impartiality-challenge

無料

Public-safe preliminary screen for challenges to the impartiality of an AI evaluator, LLM-as-a-judge, decision-support agent, or adjudicative system based on prior outputs, suspected bias, unstable reasoning, or self-reinforcing memory. Use when asked whether an AI should be recused, disqualified, replaced, audited, or escalated. Does not decide real cases or claim that human judicial-recusal doctrine automatically applies.

日本語の概要は準備中です。原文の説明を表示しています。