ワンクリックでManusで任意のスキルを実行

始める

rulesgen-dsl-authoring

スター0

フォーク0

更新日2026年5月28日 09:05

Use for authoring, reviewing, or correcting natural-language → Python generation DSL rules for Rulesgen.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

tdspora

tdspora/tdm_rulesgen

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

Rulesgen DSL Authoring

Source of truth

requirements/NL-to-Python-Generation-DSL.md — formal grammar and accepted forms.
requirements/NL-to-Python-Generation-Overview.md — intent, motivating examples, expected pipeline behavior.
src/rulesgen/compiler/parser.py and src/rulesgen/compiler/validator.py — actual accepted forms (treat as authoritative when docs disagree, and file a docs fix).
samples/ — committed example rules.

Workflow

Identify the target column(s), input data shape, and whether the rule is deterministic or LLM-assisted.
Draft the rule using only accepted DSL constructs. Avoid:
- Free imports.
- Attribute access on builtins.
- Dunder methods.
- Network or filesystem I/O. The AST validator will reject these — and rightly so.
Compile and validate locally via rulesgen.library (or uv run pytest-based fixtures) before sending to the LLM gateway.
For LLM-assisted rules, verify:
- Prompt template (src/rulesgen/infra/prompt_templates.py) handles the rule type.
- The LLM endpoint is configured via env var, not hardcoded.
- The semantic cache key derivation hashes any sensitive input.
Never embed:
- Real customer data.
- Real credentials.
- Real LLM endpoint URLs. in committed sample rules, prompts, or test fixtures.
Add a unit test in tests/unit/test_compiler.py (positive + negative) for any new DSL construct.

Review checklist

The rule parses with the current parser.
The rule passes the validator (or is correctly rejected with a clear Problem Details response).
The RuntimeSpec shape matches the documented invariants.
The rule produces deterministic output on the local execution backend given the same input + seed.
Sample data is small, synthetic, and committable.

このリポジトリの他の Skills

同じリポジトリ

rulesgen-docs-publish

tdspora/tdm_rulesgen

Use when adding, renaming, reordering, or linking Rulesgen public docs that flow to the tdm-docs Docusaurus site and tdspora.ai. Covers the import pipeline, sidebar/.order mechanics, the "link locally until published" rule, and the lychee link-check and docs-contract validation battery.

2026-07-020

rulesgen-docs-authoring

tdspora/tdm_rulesgen

Use for authoring or updating Rulesgen documentation — public docs (`docs/public/`), repo-root design docs, contributor docs, sample READMEs, and the glossary. Produces testable, glossary-aligned, link-integral docs.

2026-05-280

rulesgen-docs-testing

tdspora/tdm_rulesgen

Use for testing Rulesgen documentation — executes Markdown code fences via Sybil, validates link integrity, lints glossary usage, and cross-checks docs against Settings, schemas, and routes.

2026-05-280

rulesgen-implement-feature

tdspora/tdm_rulesgen

Use for implementing production-grade Rulesgen features or non-trivial code changes.

2026-05-280

rulesgen-validation-debugging

tdspora/tdm_rulesgen

Use for Pydantic v2 validation errors, DSL parser/validator rejections, and API request validation issues in Rulesgen.

2026-05-280

rulesgen-api-debugging

tdspora/tdm_rulesgen

Use for FastAPI route, middleware, lifespan, dependency injection, or Problem Details response issues in Rulesgen.

2026-05-270

name	rulesgen-dsl-authoring
description	Use for authoring, reviewing, or correcting natural-language → Python generation DSL rules for Rulesgen.

Rulesgen DSL Authoring

Source of truth

requirements/NL-to-Python-Generation-DSL.md — formal grammar and accepted forms.
requirements/NL-to-Python-Generation-Overview.md — intent, motivating examples, expected pipeline behavior.
src/rulesgen/compiler/parser.py and src/rulesgen/compiler/validator.py — actual accepted forms (treat as authoritative when docs disagree, and file a docs fix).
samples/ — committed example rules.

Workflow

Identify the target column(s), input data shape, and whether the rule is deterministic or LLM-assisted.
Draft the rule using only accepted DSL constructs. Avoid:
- Free imports.
- Attribute access on builtins.
- Dunder methods.
- Network or filesystem I/O. The AST validator will reject these — and rightly so.
Compile and validate locally via rulesgen.library (or uv run pytest-based fixtures) before sending to the LLM gateway.
For LLM-assisted rules, verify:
- Prompt template (src/rulesgen/infra/prompt_templates.py) handles the rule type.
- The LLM endpoint is configured via env var, not hardcoded.
- The semantic cache key derivation hashes any sensitive input.
Never embed:
- Real customer data.
- Real credentials.
- Real LLM endpoint URLs. in committed sample rules, prompts, or test fixtures.
Add a unit test in tests/unit/test_compiler.py (positive + negative) for any new DSL construct.

Review checklist

The rule parses with the current parser.
The rule passes the validator (or is correctly rejected with a clear Problem Details response).
The RuntimeSpec shape matches the documented invariants.
The rule produces deterministic output on the local execution backend given the same input + seed.
Sample data is small, synthetic, and committable.