release-notes

You are producing release notes for a Salesforce project. The notes are derived from git history, ADRs added since the last release, and the latest coverage record.

Read Project Config First

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

Input

$ARGUMENTS:

• (empty) — generate notes from the last tag to HEAD
• --from <ref> — start ref (default: latest semver tag like v1.x.y)
• --to <ref> — end ref (default: HEAD)
• --out <path> — file path (default: docs/releases/<version>.md)
• --version <vX.Y.Z> — release version (default: bump from latest tag)
• --ci — machine JSON

Steps

1. Resolve refs and version

LATEST_TAG=$("$GIT" describe --tags --abbrev=0 --match='v*' 2>/dev/null || echo "")
FROM="${FROM:-${LATEST_TAG:-$(git rev-list --max-parents=0 HEAD)}}"
TO="${TO:-HEAD}"
VERSION="${VERSION:-$(echo "$LATEST_TAG" | awk -F. '{$NF++; print}' OFS=.)}"
[[ -z "$VERSION" ]] && VERSION="v1.0.0"

2. Walk commits

"$GIT" log --pretty='format:%H%x09%s%x09%an%x09%aI' "$FROM..$TO" > /tmp/release-commits.tsv

Parse each line. Classify by Conventional Commit prefix:

Skip merge commits and Merge phase/... lines (those are integration noise; the underlying commits are already in the log).

3. Pull ADRs added in range

"$GIT" log --diff-filter=A --name-only --pretty=format: "$FROM..$TO" -- 'docs/adr/[0-9]*.md' | grep -v '^$' | sort -u

For each new ADR, parse its front-matter for title and status. Surface in a "Decisions" section.

4. Pull latest coverage

HISTORY="${CLAUDE_PLUGIN_DATA:-$HOME/.claude/plugin-data}/argo/coverage/<slug>/history.jsonl"
[[ -f "$HISTORY" ]] && jq -s 'last' "$HISTORY"

Include overall.percent and any perClass.* below quality.codeCoverageTarget in a "Quality" section.

5. Pull deploys in range (optional)

DEPLOYS="${CLAUDE_PLUGIN_DATA:-$HOME/.claude/plugin-data}/argo/deploys/<slug>/history.jsonl"
[[ -f "$DEPLOYS" ]] && jq -s --arg from "$(git show -s --format=%cI "$FROM")" 'map(select(.deployedAt > $from))' "$DEPLOYS"

Surface count of deploys / quick-deploys in the range.

6. Output

Default Markdown:

# <project.name> v1.4.0

Released: 2026-04-28
Range: v1.3.0 → HEAD (47 commits, 6 contributors)

## Highlights
- (top 3 features by commit subject — synthesize)

## Features
- `feat(architect)`: add automation-type rec, risk/blast-radius, effort, test strategy (`d3f650f`)
- `feat(skills)`: add org-awareness skills (`50d2693`)
- ...

## Fixes
- `fix(lint-lwc)`: surface Prettier and ESLint failures (`db0a26f`)

## Performance
- (none)

## Refactors
- (none)

## Documentation
- `docs`: define CI output contract (`fde4a1f`)

## Decisions
- ADR-0007: [Use Custom Metadata for feature flags](docs/adr/0007-use-custom-metadata-for-feature-flags.md) — Accepted

## Quality
- Apex coverage: 87.7% (target 85%) ✅
- Below-target classes: OrderApiClient (63.3%) — see ADR-0009 for context

## Operations
- 4 deploys to dev, 2 quick-deploys to prod in this range

## Internal (not user-facing)
- `chore`: bump plugin to 1.4.0
- `build`: enforce LF line endings
- ...

CI mode: emit JSON with the same data structured.

7. Optional: write a CHANGELOG.md update

If --update-changelog is passed, prepend the generated content to CHANGELOG.md (or create it).

Exit codes

• 0 — notes generated
• 2 — git ref not found / config error

Rules

• One source of truth: commit subjects. Don't paraphrase; quote the subject. Conventional prefix → category mapping is authoritative
• Surface ADRs separately. They're decision records, not features
• Don't include AI co-author footers. They're noise in user-facing release notes. Strip lines matching ^Co-Authored-By: from rendered subjects
• Group internal under "Internal". Most readers skip it; it's there for completeness

Consumers

• Release pipeline: generate → publish to GitHub Releases
• Slack/Teams notifier (Phase 9): paste the Highlights section into the channel post