Skip to content

AGENTS.md

Latest commit

 

History

History
74 lines (53 loc) · 3.63 KB

AGENTS.md

File metadata and controls

74 lines (53 loc) · 3.63 KB

AGENTS.md — Python Project Template

Canonical, model-agnostic agent guide. CLAUDE.md / GEMINI.md / CURSOR.md are thin wrappers that point here. Edit shared knowledge in THIS file only.

What this repo is

This repository is a Copier template repository (a meta-project). Running Copier against this repo generates a new Python project by rendering the template/ directory into a destination folder.

Warning

Copier can run template tasks during copier copy/copier update. Only use the --trust flag with templates you trust.

Behavioral guidelines

Bias toward caution over speed; use judgment on trivial tasks.

  • Think before coding — state assumptions; surface multiple interpretations instead of silently picking one; ask when unclear.
  • Simplicity first — minimum code that solves the problem; nothing speculative.
  • Surgical changes — touch only what the task requires; match existing style; don't refactor what isn't broken.
  • Goal-driven execution — define a verifiable success check before starting; loop until it passes.

Context loading (progressive disclosure)

  1. Read this file first.
  2. Load a deeper reference ONLY when the task touches its area:
When the task involves… Read
Directory structure, project layout scripts/CLAUDE.md, tests/CLAUDE.md, .github/CLAUDE.md, template/CLAUDE.md
Development setup, just recipes, CI .claude/references/dev-setup.md, .claude/references/justfile-recipes.md, .claude/references/ci-pipeline.md
Generating or updating projects .claude/references/copier-template.md
Testing strategy and patterns tests/CLAUDE.md + .claude/references/testing.md
Copier variables and Jinja2 .claude/references/copier-variables.md
Code style and standards .claude/references/code-style.md
Standards enforcement (hooks, commands, definition of done) .claude/references/standards-enforcement.md
Markdown file placement .claude/references/markdown-placement.md
File management (what not to edit) .claude/references/files-management.md
Recent improvements .claude/references/recent-improvements.md
AI rules, hooks, and skills .claude/rules/README.md, .claude/hooks/README.md
Scripts scripts/CLAUDE.md
GitHub workflows .github/CLAUDE.md
Template structure template/CLAUDE.md
  1. Do not bulk-load all references — avoid context flooding.

Memory — AgentMemory only

This repo uses AgentMemory (MCP server localhost:3111, project-scoped TEAM_ID python_project_template).

  • Session start: call memory_recall with the task/goal.
  • During work: call memory_save for durable decisions, bug root causes, conventions, gotchas (types: pattern | architecture | bug | workflow | preference).
  • Session end: call memory_save with lessons learned.
  • Do NOT create, read, or append to MEMORY.md, CONFLICTS.md, NOTES.md, or any markdown "memory" / "conflict log" file. Cross-session memory and conflict tracking live in AgentMemory exclusively.

Repo-specific guidance

Detailed how-tos live in .claude/references/ (load per the table above). Rules and enforcement hooks are under .claude/rules/ and .claude/hooks/ — referenced in .claude/rules/README.md and .claude/hooks/README.md.

AgentMemory Insights (snapshot 2026-05-21)

Architecture & Design

  • No durable insights surfaced this snapshot.

Working Preferences & Conventions

  • No durable insights surfaced this snapshot.

Bugs, Incidents & Gotchas

  • No durable insights surfaced this snapshot.

Active Projects / WIP Context

  • No durable insights surfaced this snapshot.