pr-prepare

You are preparing a pull-request body. The body is assembled from the outputs of several argo skills and presented as a single Markdown ready to paste (or push via gh).

Read Project Config First

source "${CLAUDE_PLUGIN_ROOT}/hooks/lib/config.sh"
PROJECT_NAME="$(sf_config_get '.project.name' "$ENV")"
DEFAULT_ORG="$(sf_config_get '.platform.defaultTargetOrg' "$ENV")"

Input

$ARGUMENTS:

• (empty) — generate PR body from current branch vs main
• --vs <ref> — base ref (default main)
• --out <path> — write to file (default stdout)
• --push — after generating, run gh pr create with the body (only if gh CLI is installed and authed)
• --target-org <alias> / --env <name> — for the deploy validation step
• --no-validate — skip the deploy validation step (useful when prod can't be hit)

Steps

1. Compute git context

BASE="${VS:-main}"
HEAD_REF=$("$GIT" rev-parse --abbrev-ref HEAD)
CHANGED_FILES=$("$GIT" diff --name-only "$BASE...HEAD")
COMMIT_COUNT=$("$GIT" rev-list --count "$BASE...HEAD")

If there are no changes vs the base, exit 2 with a clear message.

2. Run the skill chain in parallel (where possible)

/argo:code-review pr --ci --format json --out /tmp/pr-review.json
/argo:coverage-trend diff --vs "$BASE" --ci > /tmp/pr-coverage.json
/argo:diff-deploy --vs "$BASE" --validate --target-org "$DEFAULT_ORG" --ci > /tmp/pr-deploy.json
/argo:security-scan --ci --format json --out /tmp/pr-security.json

Capture exit codes — they become PR-body badges.

3. Synthesize summary from commits

"$GIT" log "$BASE...HEAD" --pretty='format:%s' | grep -E '^(feat|fix|perf|refactor)' | head -5

Map prefixes to bullet labels:

• feat: → "Adds"
• fix: → "Fixes"
• perf: → "Speeds up"
• refactor: → "Refactors"

4. Assemble the body

## Summary

- Adds: paginated organizations table with status filter (acmeOrgList LWC, OrgController Apex)
- Fixes: Order_Item__c master-detail cascade delete bug (OrderTriggerHandler)
- Refactors: extracted shared retry policy into RetryPolicy.cls

(Drawn from <COMMIT_COUNT> commits since `<BASE>`.)

## Test plan

- [x] LWC unit tests pass (Jest, 47 tests, 0 failures)
- [x] Apex unit tests pass (run via /argo:test-coverage)
- [x] Apex coverage at or above target (87.7% overall, target 85%)
- [x] /argo:code-review pr — no critical findings
- [x] /argo:diff-deploy --validate against DevSandbox succeeded
- [x] /argo:security-scan — no critical findings
- [ ] Manual smoke (browser) — _reviewer to verify the LWC renders against acmeOrgList page_

## Gates

| Gate | Status |
|------|--------|
| Code review (PR scope) | ✅ no critical |
| Coverage trend         | ✅ +0.2% overall, OrderApiClient regressed -8.7% (still > target) |
| Deploy validation      | ✅ 12 components, tests passed |
| Security scan          | ⚠️ 1 medium (SEC-USE-NAMED-CRED in legacy file — pre-existing) |

## Components changed

- LightningComponentBundle: acmeOrgList (new), acmeOrgRow (new)
- ApexClass: OrgController (new), OrgService (new), OrderTriggerHandler (modified)
- CustomObject (field-only changes): Order_Item__c.Discount_Pct__c (new)

## ADRs

(no new ADRs in this PR)

## Risk

- **Risk level**: Medium (master-detail relationship change on Order_Item__c)
- **Reversibility**: Backfill query + destructive change available; documented in ADR-0011
- **Mitigation**: validate against sandbox first, deploy off-hours

🤖 Body generated by /argo:pr-prepare. Re-run after pushing changes.

If --push:

gh pr create --title "$(git log -1 --format=%s)" --body "$(cat /tmp/pr-body.md)"

The title comes from the latest commit subject by default; user can override.

5. Output

Stdout (or --out file): the assembled Markdown. CI mode: same body, plus exit 1 if any gate failed.

Exit codes

• 0 — body generated, all gates pass
• 1 — body generated, at least one gate failed (caller decides whether to push anyway)
• 2 — invocation error (no diff vs base, sub-skill failures)

Rules

• Always include the Test plan as an unchecked list for the reviewer's manual smoke step. Automation can check most rows; the reviewer keeps the last one
• Don't auto-push without --push. Generating ≠ pushing; the user reviews first
• Cite sub-skill outputs. If /code-review flagged a critical, the PR body shows it in Summary so the reviewer doesn't have to dig
• Don't include the AI co-author footer in the PR body. Strip lines matching ^Co-Authored-By: from the synthesized commit-subject summary
• Skip "ADRs" section if none added. Don't pad the body with empty sections

Consumers

• Devs run before opening a PR
• CI runs to gate PR merges (looks for the body to be present and gates to pass)