Skip to content

AGENTS.md

Latest commit

 

History

History
51 lines (38 loc) · 2.45 KB

AGENTS.md

File metadata and controls

51 lines (38 loc) · 2.45 KB

Code-Mint

Code-mint is the source repository for harness-engineering documentation, skills, and templates (portable reference material for target repos). It is not an executable product. Do not attempt to build, test, or run this repository.

Editing Priorities

When working in this repo:

  1. Keep README.md as the single human-facing entry point.
  2. Keep this file agent-facing and concise.
  3. Preserve stable skill IDs unless there is a strong compatibility reason to rename them.
  4. Treat docs/onboarding-checklist.md as the canonical progress tracker and docs/outcomes.md as the outcome glossary.
  5. Update cross-links when adding or moving onboarding docs so the repo map stays trustworthy.

Repository Map

Path Purpose
README.md Human-facing overview, quick start, and onboarding journey
AGENTS.md Concise agent-facing instructions for this repository (this file)
.agents/skills/ Portable skills copied into target repositories
.agents/rules/ Guidance on project-specific persistent rules
.agents/code-mint-status.json Machine-readable onboarding fingerprint (committed)
docs/framework.md Founding principles and mechanical standards
docs/outcomes.md North-star onboarding outcomes and evidence model
docs/onboarding-checklist.md Canonical outcome checklist template
docs/adoption-guide.md Detailed operator workflow for onboarding a target repo
docs/skills-status.md Compatibility matrix from skills to outcomes
docs/skill-development.md How to create and maintain skills
.agents/skills/legibility--enhancer/references/ Root and subdirectory AGENTS.md templates

Skill Categories

Category Meaning
meta-- Onboarding and skill-library management
legibility-- Repo mapping and AGENTS.md creation
autonomy-- Self-verification, runtime readiness, env setup, and operational tooling (including autonomy--sre-*)
clarity-- Collaborative planning and executable tickets (post-core onboarding)

Working Model

Code-mint is intentionally outcome-first:

  • the user should quickly understand what the onboarding will unlock
  • progress should be visible in a checklist the user can revisit later
  • proof should be recorded as reports, smoke paths, failing tests, or investigation notes

Standards

Use the exact filename AGENTS.md. Subdirectory AGENTS.md files override parent files rather than merging with them.