wortschatz

Imported from thom404/wortschatz (AGENTS.md). Install upstream with npx skills add thom404/wortschatz. Copyright stays with the author.

wortschatz

Autonomous issue agent

Adding the agent-run label to a fully specified issue starts .github/workflows/issue-agent.yml, which runs the headless OpenCode agent .opencode/agent/issue-implementer.md. ready-for-agent is a marker only: it records that the issue is specified and starts nothing. The workflow shows agent-running while a run is in flight and clears it when the run ends. A run whose issue still has open same-repo blockers (GitHub's native dependencies) is skipped before the agent starts: it removes agent-run, comments the blocking issues, and asks the reader to add agent-run again once they close. Otherwise the agent edits the working tree only; the workflow commits, pushes feature/<n>-<slug>, and opens a pull request that says Closes #<n>. The agent never merges or closes anything, and a failed run comments the failure and applies ready-for-human. See ADR-0006. Secret setup is in docs/OPERATOR-STEPS.md §8: OPENCODE_API_KEY and a fine-grained AGENT_GITHUB_TOKEN (contents, pull-request, and issue write).

German UI copy

All user-facing strings are German and hardcoded inline. Write new copy in the recorded voice and follow the preferred/avoided terms in docs/ux-copy.md; the decision is ADR-0007.

Visual regression tests

Vitest + happy-dom stays the fast behavior layer (npm test). A separate Playwright layer renders the production client in Chromium and compares full-page screenshots against committed baselines; it never runs as part of npm test, the Docker build or the issue agent. See ADR-0020.

• npm run test:visual — build and serve the production client, then run the 66-row screenshot matrix, six ARIA snapshots and five flow tests.

• npm run test:visual:update — accept intended visual changes. Run this only for a deliberate UI change and review the changed PNGs in the pull request; never run it just to make a red job green.

• Baselines live in e2e/__screenshots__/ (PNG) and e2e/__aria__/ (YAML) and are committed. Update and verify them inside the pinned container so local and CI pixels match:

  docker run --rm --ipc=host --user "$(id -u):$(id -g)" -e HOME=/tmp \
    -v "$PWD":/work -w /work \
    mcr.microsoft.com/playwright:v1.64.1-noble \
    bash -lc 'npm ci && npm run test:visual:update'

  Use npm run test:visual in the same image to verify without updating. The image tag must match @playwright/test in package.json.

• The suite freezes the page clock, pins de-DE / Europe/Berlin and reducedMotion, disables animations, uses no masks, and fails on any network request outside the preview origin. A flaky screenshot is a fixture or app defect: fix it at the source rather than adding a retry or mask.

• The visual CI job is additive and required; workflow files are human-owned (ADR-0006 §5), so adding it is a human step.

Agent skills

Issue tracker

Issues and specs live as GitHub issues in thom404/wortschatz, via the gh CLI. See docs/agents/issue-tracker.md.

Triage labels

Default five-role vocabulary (needs-triage, needs-info, ready-for-agent, ready-for-human, wontfix). See docs/agents/triage-labels.md.

Domain docs

Single-context: one GLOSSARY.md + docs/adr/ at the repo root. See docs/agents/domain.md.

Worktree sessions

docs/agents/worktrees.md has the create, dependency, check and cleanup recipe.