Skip to content

CLAUDE.md

Latest commit

 

History

History
52 lines (40 loc) · 2.42 KB

CLAUDE.md

File metadata and controls

52 lines (40 loc) · 2.42 KB
Raw

CLAUDE.md

Context for Claude Code in this repo. Keep this short — it loads on every session.

What this is

claude-code-commons is a library, not a routine. It holds the small, vendor-agnostic, standard-library-only helpers shared by Claude routine repos: env_required/env_opt/env_get, env_int/env_float, parse_num, currency_symbol, build_remote_path, log. It has no vendor or report logic and no third-party dependency.

There is no scheduled-task/operator section here — this package does not run on its own. It is consumed by the routine repos and by vendor toolkits (e.g. claude-woocommerce-commons), which own their own runtime contracts.

Read first

Read SCOPE.md before changing anything. Every feature has an ID (F-001…) and must stay covered by a test (tests/test_scope_coverage.py enforces it).

Working rules

  • Any behaviour change starts in SCOPE.md: add/edit a feature ID, then write or update its test, then change the code. Never the other way around.
  • Every test is tagged @pytest.mark.feature("F-00X").
  • Run pytest after any change.
  • After a behaviour change, add a line under [Unreleased] in CHANGELOG.md.
  • Before committing, review: adversarially re-read the full uncommitted diff for real defects — edge cases, wrong assumptions, criteria the tests miss (in Claude Code, /code-review). Style is not a finding (ruff owns style); a session habit, not a CI gate.
  • No third-party dependency, ever. The whole point is that a routine can install this without pulling in heavier libraries. Anything needing requests, openpyxl, etc. belongs in a vendor toolkit that depends on this, not here.
  • This is a dependency of routines and vendor toolkits. A breaking change to a public name must bump the version and be rolled out to each consumer by bumping its pin. Prefer additive changes.

Deployment floor

requires-python is >=3.9 and CI runs a 3.9 / 3.11 / 3.12 matrix. The Cowork sandbox that runs scheduled routines is Python 3.11, so a consumer that pip installs this must be able to; keeping the floor at the lowest interpreter any routine runs on (and testing it) guarantees that.

Commands

  • Install (editable, with tests): pip install -e ".[test]"
  • Run tests: pytest
  • The package exposes a single top-level module: from code_commons import env_required, env_opt, parse_num, build_remote_path, currency_symbol, log