create-pr

Prompt file imported from RoboFinSystems/xbrlkit-viewer (.claude/commands/create-pr.md). Fill in {{arguments}} before use. Copyright stays with the author.

Create a GitHub pull request for the current branch, writing the title and description from the actual work done in this session — not reconstructed from the diff.

Why this command exists

A description written from the diff alone can't know why a change was made, so it tends to describe things that aren't true. You author it here, where the full context is available.

This is the xbrlkit viewer — a static, client-side React/Vite renderer for holon.jsonld and tavi.json reports, deployed as a bundle behind CloudFront. There is no backend, and users supply their own Anthropic, ElevenLabs, and RoboSystems API keys in the browser. This repository is public.

Instructions

1. Preflight

CURRENT=$(git branch --show-current)
TARGET=${1:-main}

• Never PR from main. Branches come from npm run feature:create, not git switch -c.
• Source ≠ target. Stop if equal.
• Uncommitted changes: surface them and ask whether to commit (never on main, stage by name, no git add -A).
• Existing PR: gh pr list --head "$CURRENT" --base "$TARGET" --json url,number — offer gh pr edit rather than duplicating.
• Push: git push -u origin "$CURRENT".

2. Gather the real change context

Use this session as the primary source; corroborate with git log, git diff --stat, and the full git diff. No confabulation — every claim must be supported by the diff.

3. Compose the PR

• Title — conventional-commit style with a scope, matching git log.

• Body — match the headings in .github/PULL_REQUEST_TEMPLATE.md, since --body-file bypasses template prefill and silently drops omitted sections:

  • Summary — 1–3 sentences.
  • Changes — grouped by area: report rendering, modes (file/SEC), keys drawer, AI + SPARQL, voice, app shell.
  • Key Handling — the section that matters most here. "No change" if the diff doesn't touch key storage, transport, or any outbound request. Otherwise name every destination the change can send data to, and confirm each is a provider the user knowingly supplied a key for.
  • Testing — the gate is npm run test:all (format:check → lint → typecheck → test → build), and it's check-only. Say whether you actually loaded a report — unit tests don't cover rendering, and file mode's offline guarantee isn't asserted anywhere.

  Put Closes #123 as the last line of the Summary.

• Say if rendering output changed. A different value, subtotal, or dimensional breakdown is what users read as a financial statement. Note that most statement rendering lives in @robosystems/report-components — if the fix belongs there, say so rather than patching locally, since a local patch is overwritten on the next bump.

• Say if file mode's offline guarantee is affected. Opening a local holon must make no network call. Any new fetch on that path is a stated-promise break, not an optimization detail.

• Attribution — attribute to the user only; no Claude footer or trailer unless explicitly asked.

4. Create the PR

gh pr create --base "$TARGET" --head "$CURRENT" --title "<title>" --body-file /tmp/pr-body.md

5. Optional Claude review

Only if the user asks (review / --review): gh pr comment <number> --body "@claude please review this PR".

Output

1. PR URL. 2. Title. 3. Target ← source. 4. Whether key handling or the offline guarantee is affected. 5. Whether a review was requested.

{{arguments}}