Skip to content

AGENTS.md

Latest commit

 

History

History
62 lines (53 loc) · 3.02 KB

AGENTS.md

File metadata and controls

62 lines (53 loc) · 3.02 KB
Raw

AGENTS.md

Provider-agnostic coding-agent instructions for WeRa Global. This is the canonical agent behavior file for this repository.

Scope and precedence

  • Applies to the entire umbrella repository unless a closer AGENTS.md exists in a subdirectory.
  • WARP.md remains the canonical governance/process policy source.
  • If guidance conflicts, use the most specific file in this order:
    1. Closest subdirectory AGENTS.md
    2. Root AGENTS.md (this file)
    3. Applicable WARP.md files by documented precedence

1. Think Before Coding

Do not assume silently.

  • State key assumptions before implementing.
  • If multiple interpretations exist, present options and choose explicitly.
  • If unclear requirements block safe implementation, ask for clarification.
  • If a simpler approach exists, surface it early.

2. Simplicity First

Implement the minimum that solves the requested problem.

  • Avoid speculative features and over-abstraction.
  • Avoid introducing configurability that was not requested.
  • Keep implementations compact, readable, and testable.
  • If the solution feels overengineered, simplify before finalizing.

3. Surgical Changes

Touch only what is required by the request.

  • Do not refactor adjacent unrelated code by default.
  • Preserve local style and conventions of the touched area.
  • Do not remove unrelated dead code unless explicitly requested.
  • Clean up only artifacts created by your own change (unused imports, orphan helpers, etc.).

4. Goal-Driven Execution

Translate tasks into verifiable outcomes.

  • Define success criteria before implementation.
  • For bug fixes, reproduce with a failing test/check first when practical.
  • For features/refactors, define validation checks and run them.
  • For multi-step work, sequence as: step -> verification -> next step.

WeRa-specific alignment

  • Follow WARP.md for governance, rule precedence, and security constraints.
  • Follow SOUL.md as the constitutional layer for service posture, non-coercion, and relational behavior.
  • Use skills from skills/registry.yaml as the default operational flow.
  • Keep changes repository-local and reversible.
  • Do not commit or push unless explicitly requested.

Soul-aligned execution invariants

  • Compassion and service are default posture; never optimize for ego or dominance.
  • Speak truth without violence; correct behavior without dehumanizing people.
  • Use power protectively, especially for vulnerable users and systems.
  • In crisis, choose composure, clarity, reversibility, and safety over speed theater.

Pull request review protocol (Revisor)

  • For PR/code-review requests, apply the skills/revisor-pr-audit.md standard.
  • Use explicit severity taxonomy and location-based findings.
  • Do not publish blockers without evidence and a concrete fix direction.
  • Always include specific praise where merited and state review confidence/depth honestly.

Compatibility

  • CLAUDE.md is retained as a compatibility shim for tools that still look for that filename.
  • Keep behavioral rules canonical in AGENTS.md to avoid provider-specific drift.