360 Secure SkillHub
360 安全技能中心

engineering-rules

提供按需读取的团队工程规则集,涵盖 Python 工具链、包布局、深模块架构、测试、日志,以及 FastAPI、Prefect、LLM、ASR、Next.js、k8s 等栈规范,用于代码编写、评审或重构时查规则并引用规则 id。

collectiveai-team
热度 0.00
最近更新 2026-10-19 07:45:34
分类:开发依赖:读写文件应用场景:代码审查 / 代码生成 / API 集成 / DevOps

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add collectiveai-team/botica/engineering-rules
Metadata
name
engineering-rules
metadata
{"credits":[{"author":"Matt Pocock","skill":"codebase-design","url":"https://github.com/mattpocock/skills/blob/v1.3.1/skills/engineering/codebase-design/SKILL.md","contribution":"Deep-module vocabulary, deletion test, adapter discipline, and interface-based testing; other house rules are local."}]}
description
The house engineering ruleset — Python toolchain, package layout, deep-module architecture, testing, logging, and the stack-specific patterns (FastAPI, Prefect, LLM/AI, ASR, Next.js, k8s). Use when looking up what the house standard says about code being written, reviewed, refactored, or restructured — which linter, which layout, which test shape, whether a pattern is current or legacy. Invoked by engineering-pr-review and engineering-refactor, and usable directly.

Engineering Rules

The single source of the house engineering rules. This skill owns the ruleset; the workflows that apply it (engineering-pr-review, engineering-refactor) invoke this one rather than carrying their own copy.

Reading protocol

Read by need, not all at once. The ruleset is ~3,800 lines; loading it whole wastes context that the actual work needs.

  1. rules/_sections.md — the index. Every rule id, grouped by section, one line each. Start here always.
  2. rules/<id>.md — the specific rules the work touches. Read only these.
  3. AGENTS.md — the whole ruleset compiled into one document. For skimming everything at once; not the default path.

Rule ids are kebab-case and section-prefixed (py-package-manager is in section py), so the index tells you which files to open from the file paths alone.

Judging a rule before applying it

Each rule carries two fields that decide whether it applies here at all:

  • applies-to — all, or a stack tag (python, fastapi, prefect, asr, llm, nextjs, k8s, edge). A Prefect rule says nothing about a repo with no Prefect.
  • status — current (what the repos do today), direction (where new or modernized code should head), legacy (preserve where it exists; don't expand it, don't fight it).

The local repo wins over the direction unless the user explicitly asked to modernize (general-respect-local-repo). A repo's package manager, lockfile, CI, configured linter, and public interfaces are contracts — judge code against what its repo actually uses, not against a one-size-fits-all standard.

Citing rules

Cite the rule id, not a file path: "violates py-no-formatter-churn", not "violates rules/py-no-formatter-churn.md". The id is the stable name; the path is an implementation detail of this bundle.

Relationship to the CES catalog

In a repo scaffolded by collectiveai-team/botica, the house standards also ship as always-on context: an index in AGENTS.md plus detail files under .agents/rules/<slug>.md, each carrying a CES-<issue#> code (see docs/engineering-standards.md).

Where both are present, .agents/rules/ is authoritative — it is the tracked, coded, enforced catalog, and several rule ids here share its slugs. This bundle exists to carry the rules into repos that have no .agents/rules/ at all.