policyengine-standards
PolicyEngine coding standards, environment setup, formatters, changelog, and CI requirements across every repo — what makes a PR pass and merge cleanly. Covers uv + .venv, ruff, bun, towncrier changelog fragments, the "wait for CI" PR pattern, and the repo-rename checklist. Triggers: "CI failing", "linting", "formatting", "before committing", "PR standards", "code style", "ruff", "prettier", "changelog", "towncrier", "create a PR", "wait for CI", "rename a repo", "set up the venv", "which Python version".
PolicyEngine · Static checks on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add PolicyEngine/policyengine-claude --skill policyengine-standardsAbout this skill
name: policyengine-standards description: | PolicyEngine coding standards, environment setup, formatters, changelog, and CI requirements across every repo — what makes a PR pass and merge cleanly. Covers uv + .venv, ruff, bun, towncrier changelog fragments, the "wait for CI" PR pattern, and the repo-rename checklist. Triggers: "CI failing", "linting", "formatting", "before committing", "PR standards", "code style", "ruff", "prettier", "changelog", "towncrier", "create a PR", "wait for CI", "rename a repo", "set up the venv", "which Python version". metadata: category: process
PolicyEngine development standards
Load this before committing to, reviewing, or setting up any PolicyEngine repository. It encodes the org conventions that CI enforces — the things a frontier model would otherwise guess wrong. For prose style (blog posts, PR bodies, docs) see the policyengine-writing skill.
Environment: uv + a local .venv
All PolicyEngine Python work uses uv with a .venv at the
repository root. Never use bare pip, and never run bare python/pytest.
uv venv # creates ./.venv (uv picks the interpreter from pyproject.toml)
uv pip install -e ".[dev]" # editable install with dev extras
uv run pytest # run anything through uv so it uses ./.venv
- Python version is per-repo — check
pyproject.toml(requires-python). The org spans the 3.13/3.14 era; there is no single global version. Do not downgrade a repo to satisfy a stale assumption, and do not hardcode a version across repos. uv run <cmd>guarantees the command runs in the repo's.venvwith its locked deps. Barepython script.py/pytestis the most common cause of "works locally, fails in CI".- Verifying "latest": if asked for the latest version of a package, read PyPI
(
https://pypi.org/pypi/<pkg>/json→info.version) immediately before pinning. Never infer "latest" from a lockfile, a local install, or search snippets.
Documentation builds
Python docs use Jupyter Book 2.0 (MyST-NB) — build with myst build docs, never jb build
(that is Jupyter Book 1.x). Use MyST markdown syntax.
Before committing — checklist
- Write the test first (TDD, below).
- Format:
make format(oruv run ruff format .for Python,bun run lint -- --fix && bunx prettier --write .for JS). - Run tests:
make test(oruv run pytest). - Add a changelog fragment (towncrier, below).
- Reference the issue in the PR: "Fixes #123".
Creating pull requests
The CI waiting problem
Common failure pattern:
User: "Create a PR and mark it ready when CI passes"
Claude: "I've created the PR as draft. CI will take a while, I'll check back later..."
[Chat ends - Claude never checks back]
Result: PR stays in draft, user has to manually check CI and mark ready
Solution: use the /create-pr command
When creating PRs, use the /create-pr command. It creates the PR as draft, actually waits
for CI (polls every 15 seconds), marks it ready when CI passes, and reports failures with
details. It works because the command carries explicit polling logic that Claude executes, so
it waits instead of giving up.
If /create-pr is not available
Implement the polling pattern directly:
# 1. Create PR as draft. Create the branch ON the PolicyEngine repo, not a fork.
gh pr create --repo PolicyEngine/policyengine-us --draft --title "Title" --body "Body"
PR_NUMBER=$(gh pr view --json number --jq '.number')
# 2. Wait for CI (ACTUALLY WAIT - don't give up!)
POLL_INTERVAL=15
ELAPSED=0
while true; do # No timeout - wait as long as needed
CHECKS=$(gh pr checks $PR_NUMBER --json status,conclusion)
TOTAL=$(echo "$CHECKS" | jq '. | length')
COMPLETED=$(echo "$CHECKS" | jq '[.[] | select(.status == "COMPLETED")] | length')
echo "[$ELAPSED s] CI: $COMPLETED/$TOTAL completed"
if [ "$COMPLETED" -eq "$TOTAL" ] && [ "$TOTAL" -gt 0 ]; then
FAILED=$(echo "$CHECKS" | jq '[.[] | select(.conclusion == "FAILURE")] | length')
if [ "$FAILED" -eq 0 ]; then
echo "All CI passed. Marking ready..."
gh pr ready $PR_NUMBER
break
else
echo "CI failed. PR remains draft."
gh pr checks $PR_NUMBER
break
fi
fi
sleep $POLL_INTERVAL
ELAPSED=$((ELAPSED + POLL_INTERVAL))
done
Never say "I'll check back later" — the chat session ends. Default to draft PRs; mark ready
only when the user asks or CI is already verified green. Merge only after every check passes
and the PR is MERGEABLE.
Test-driven development
PolicyEngine follows TDD: write the test (RED), implement (GREEN), refactor. In multi-agent encoding workflows, test authoring and rule implementation proceed independently from the same regulations so the test is not shaped to the implementation.
Country models test with YAML integration cases, not hand-built Python situations. A case declares inputs and expected outputs against named variables; the runner builds the simulation:
# policyengine_us/tests/.../ctc.yaml
- name: CTC for a married couple with two children
period: 2026
input:
people:
parent1: {age: 40, employment_income: 75_000}
parent2: {age: 40, employment_income: 50_000}
child1: {age: 5, is_tax_unit_dependent: true}
child2: {age: 8, is_tax_unit_dependent: true}
tax_units:
tax_unit:
members: [parent1, parent2, child1, child2]
filing_status: JOINT
output:
ctc: 4_400 # $2,200 per child under post-OBBBA law in 2026
Run country-model YAML tests through the core test runner (this is what make test wraps):
uv run policyengine-core test policyengine_us/tests/policy/ -c policyengine_us
For non-model Python packages, write pytest tests under tests/ mirroring the source layout;
run with uv run pytest tests/ -v. Test behavior, not implementation; give tests clear names
and docstrings citin