Skip to content

dd3ok/boring-backend

Repository files navigation

Boring Backend Skill

한국어

Boring Backend is a compact skill for AI coding agents that design, implement, or review API/service reliability work.

Design Bias

Boring Backend is built for AI coding agents that can over-design, miss edge cases, or overstate confidence. It intentionally blends:

  • Test-aware problem framing: start from failure modes, not happy paths. Every guard should end in runnable evidence when possible. Static review and checklists help, but missing tests or smoke runs lower confidence.
  • Agent hygiene: keep changes surgical, name assumptions, choose the smallest working path, and define success in commands the agent can run.
  • SOLID + YAGNI balance: separate real responsibilities such as routing, domain rules, persistence, DTOs, and error mapping only when the current contract needs it. Avoid interfaces, factories, strategies, or plugin layers for future variants.

The intended advantage is one clear trigger with internal modes for design, implementation, review, and production-evidence runs. That keeps discovery simple while still protecting correctness, security, data integrity, status codes, performance, and operational guardrails.

Skill

  • boring-backend: design, implement, or review API/service reliability for auth, data integrity, idempotency, concurrency, performance, distributed behavior, and operational risk.

The skill uses one trigger with four internal modes:

  • Design: define contracts, invariants, guard strategy, tradeoffs, and evidence targets before implementation.
  • Implementation: make scoped API/service changes with tests and guard evidence.
  • Review: check code for P0-P4 reliability, security, data integrity, performance, compatibility, and operations risks.
  • Production evidence: separate local evidence from load, query-plan, latency, saturation, rollout, rollback, and observability evidence.

Layout

  • skills/boring-backend/: source skill package.
  • .agents/skills/boring-backend/: project-local Codex/Antigravity-style mirror.
  • .claude/skills/boring-backend/: project-local Claude Code mirror.
  • validation/: repository-level behavior, trigger, and fairness evaluation inputs; intentionally outside the installed runtime skill.
  • scripts/verify_all.py: runs mirror and repository checks.
  • scripts/verify_boring_backend_skill_mirrors.py: verifies source and mirror packages stay in sync.
  • reports/: ignored workspace for generated evaluation output; retained only as an empty local target.

Install

With Codex skill-installer, install only the runtime skill folder:

--repo dd3ok/boring-backend --path skills/boring-backend

Manual install is also path-only: copy skills/boring-backend into your runtime's skills directory.

Common local targets:

Runtime Project scope User scope
Codex / Agents .agents/skills/boring-backend $HOME/.agents/skills/boring-backend
Claude Code .claude/skills/boring-backend ~/.claude/skills/boring-backend
Antigravity .agents/skills/boring-backend ~/.gemini/config/skills/boring-backend

Do not install .agents/, .claude/, validation/, reports/, or scripts/ as runtime skills. They are development mirrors, evaluation assets, generated output, and verification utilities.

Verification

Install the development dependency in a project-local Python 3 virtual environment:

python -m pip install -r requirements-dev.txt

Then run the verification entrypoint from the repository root with the active environment or the platform's Python 3 launcher:

python scripts/verify_all.py
python3 scripts/verify_all.py  # macOS/Linux
py -3 scripts/verify_all.py    # Windows

GitHub Actions runs the same entrypoint on Ubuntu, macOS, and Windows.

License

MIT. See LICENSE.

About

A compact AI coding-agent skill for designing, implementing, and reviewing boringly reliable backend APIs.

Topics

Resources

License

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages