Skip to content
This repository was archived by the owner on Apr 20, 2026. It is now read-only.

Latest commit

 

History

History
99 lines (74 loc) · 4.35 KB

File metadata and controls

99 lines (74 loc) · 4.35 KB

AGENTS.md - Local Configuration

Hierarchy

Remote AGENTS.md at https://raw.githubusercontent.com/joaomj/skills/main/AGENTS.md is the primary source of truth.

Core Principles

Principle Description
sdd-first Spec-Driven Design: specs before tests, tests before implementation. Bug fixes require spec + regression test.
no-hardcoding All configurable values in config module using pydantic-settings.
no-test-skipping Never use # noqa, @pytest.mark.skip, @pytest.mark.xfail in tests. Fix root cause.
intent-driven Match user intent, not exact phrases. LLMs understand synonyms and indirect requests — act on meaning, not keyword triggers.

Intent-Driven Agent Routing

When the user's intent matches a category below, invoke the corresponding agent or load the skill. Do not require exact phrase matches.

Intent Category Action
Code review / feedback on code quality @code-reviewer
Documentation cleanup / accuracy / pruning @doc-maintainer
Code simplification / reducing complexity @simplifier
CI/CD pipelines / GitHub Actions / workflows Load github-cicd-lite or docker-best-practices skill as appropriate
Pull request creation Load create-pull-request skill
Web scraping / extracting content from URLs Load firecrawl-web-scraper skill
Jira issue management Load jira-issues skill
Notion search / fetch notion docs / read notion page Load notion-reader skill
Planning a feature or implementation Switch to @plan agent (or press Tab to select Plan)
Bug fixing Write a regression test that reproduces the bug before fixing
Architecture / infrastructure / system diagram Load architecture-diagram skill

Context-Aware Skill Loading

Load relevant skills when the context matches, regardless of whether it's a file pattern, import, or conversation topic.

Domain Skills

Context Skill
Docker / containers / Dockerfile / docker-compose docker-best-practices
Machine learning / training / model pipelines / pandas / numpy / sklearn / torch ml-best-practices
Python project setup / pydantic / pytest / fastapi / flask / django python-best-practices
New or unfamiliar library import (import X) context7 (fetch up-to-date docs)
Notion / notion API / fetch notion content notion-reader
Architecture diagram / system diagram / infrastructure visualization architecture-diagram
.env.example file STOP — see env-files rule (never read .env)

When a skill fetch fails, ask the user: "Proceed without docs?"

Skill Loading Protocol

  1. Ask the user before loading a skill (unless they directly requested it).
  2. Example: when opening a Dockerfile, ask "Load Docker best practices?" before invoking the skill.
  3. If the user's request already implies the domain (e.g., "help me write a Dockerfile"), load without asking.

Workflow Triggers

Condition Action
After code changes that affect public API or user-facing behavior Suggest documentation updates
Phase gate passed Ask whether to commit
User requests commit Commit changes (no push unless requested)
Task completed Ask: "Update daily activity log?"

NON-NEGOTIABLE RULES

Rule ID Domain Rule
OC001 Type Safety No raw dicts for API schemas — use Pydantic models
OC002 Security Never view .env content (use .env.example for schema only)
OC003 Security No privileged containers
OC004 Grep-ability Absolute imports preferred over relative
OC005 Type Safety Strict type hints for all functions
OC006 Process SDD: specs before tests, tests before implementation
OC007 Quality 80% test coverage minimum
OC008 Test Integrity Zero # noqa or skip in tests — fix root cause
OC009 Supply Chain Lockfile required and committed
OC010 Supply Chain Delayed ingestion with 7-day buffer recommended

SUBAGENT INDEX

Subagent Invoke Description
code-reviewer @code-reviewer Expert code review with P0-P3 severity
simplifier @simplifier Apply project standards to simplify code
doc-maintainer @doc-maintainer Update and prune documentation