/release-sync

description: "Syncs latest release content to NotebookLM and HQ Knowledge Base after version tagging. Reads CHANGELOG, CLAUDE.md, and hook README, updates notebook sources, and ingests release digest. Optionally generates podcast from updated knowledge base. Use after tagging a new version to propagate release knowledge."
argument-hint: "[version]"
context: inherit
agent: release-engineer
user-invocable: true
name: release-sync
allowed-tools: [Bash, Read, Write, Edit, Glob, Grep]

Auto-generated from skills/release-sync/SKILL.md

Source: https://github.com/yonatangross/orchestkit

Release Content Sync

Sync the latest OrchestKit release to external knowledge systems.

> **CC ≥ 2.1.118 (M122):** Sync triggers on the new `claude plugin tag` annotated tag (in addition to plain `git tag`). The tag's annotation embeds the plugin manifest version, which release-sync uses as the canonical source-of-truth for the version being synced. See `src/skills/chain-patterns/references/plugin-tag.md`.

What This Does

1. Reads the latest CHANGELOG entry, CLAUDE.md, and hook README 2. Updates the OrchestKit NotebookLM KB notebook with fresh sources 3. Ingests the release digest into HQ Knowledge Base (if available) 4. Optionally generates a new podcast from the updated notebook

Prerequisites

• MCP servers: `notebooklm-mcp` and/or `hq-content`
• NotebookLM notebook ID stored in `.claude/release-sync-config.json`

Step 1: Detect Version and Read Sources

# Read current version from CLAUDE.md
version = Grep(pattern="Current.*\\d+\\.\\d+\\.\\d+", path="CLAUDE.md")

# Read CHANGELOG — extract latest release section
changelog = Read("CHANGELOG.md", limit=80)

# Read hook architecture summary
hook_readme = Read("src/hooks/README.md", limit=100)

# Read CLAUDE.md for project overview
claude_md = Read("CLAUDE.md")

Step 2: Load Config

config = Read(".claude/release-sync-config.json")
# Expected format:
# {
#   "notebooklm_notebook_id": "0a05e680-e33b-4d8c-94b2-5df26a1af329",
#   "hq_kb_project": "orchestkit"
# }

If config doesn't exist, prompt user:

AskUserQuestion(questions=[{
  "question": "NotebookLM notebook ID for OrchestKit KB?",
  "header": "Configuration",
  "options": [
    {"label": "Use default", "description": "OrchestKit v7 — Complete KB (0a05e680...)"},
    {"label": "I'll provide", "description": "Enter a custom notebook ID"}
  ]
}])

Step 2b: Choose Sync Targets via ork-elicit (M118 #1468)

Replace the historical 3-question sequential ask with one form using the `release-sync-targets` ork-elicit preset (registered in `src/mcp-server/src/presets/release-sync-targets.ts`). Form fields: `notebooklm`, `hq_kb`, `slack`, `notes`.

# Skip the form when targets are explicit:
#   /ork:release-sync --targets=notebooklm,slack  → skip, use those
#
# Otherwise, prefer ork-elicit; fall back to AskUserQuestion if MCP unavailable:

elicit_available = ToolSearch(query="select:mcp__ork-elicit__ork_elicit").found

if elicit_available:
    raw = mcp__ork-elicit__ork_elicit(preset="release-sync-targets")
    parsed = json.loads(raw)
    # parsed = {
    #   "action": "accept" | "decline" | "cancel",
    #   "values": {"notebooklm": bool, "hq_kb": bool, "slack": bool, "notes": str}
    # }
    if parsed["action"] != "accept":
        return  # user cancelled
    targets = parsed["values"]
else:
    # Fallback: 3 sequential AskUserQuestion calls (one per boolean target).
    # ALL targets default to False — OrchestKit is open-source and these
    # targets (notebook IDs, HQ KB, Slack) are user-private infrastructure
    # that the plugin cannot assume is configured. User explicitly opts in.
    targets = {
      "notebooklm": ask_yn("Push to NotebookLM?", default=False),
      "hq_kb":      ask_yn("Push to HQ KB?",       default=False),
      "slack":      ask_yn("Announce in Slack?",   default=False),
      "notes":      ""
    }

The form path is ~1 round-trip; the AskUserQuestion fallback is 3. Behavior at the dispatch layer is identical — the rest of this skill reads `targets["notebooklm"]`, `targets["hq_kb"]`, `targets["slack"]`, `targets["notes"]` regardless of source.

**Privacy note:** all targets are opt-in (default false). NotebookLM notebook IDs and HQ knowledge bases are user-private — the open-source plugin must not assume they exist or push to them implicitly.

Step 3: Update NotebookLM Sources

# Probe MCP availability
ToolSearch(query="select:mcp__notebooklm-mcp__source_add")

# Add release digest as new source
mcp__notebooklm-mcp__source_add(
  notebook_id=config.notebooklm_notebook_id,
  source_type="text",
  title=f"Release {version} — {date}",
  text=release_digest,
  wait=True
)

Step 4: Ingest to HQ Knowledge Base (Optional)

# Probe HQ content MCP
ToolSearch(query="select:mcp__hq-content__knowledge_ingest")

# If available, ingest
mcp__hq-content__knowledge_ingest(
  title=f"OrchestKit {version} Release Notes",
  content=release_digest,
  project="orchestkit",
  content_type="release-notes"
)

Step 5: Generate Podcast (Optional)

AskUserQuestion(questions=[{
  "question": "Generate a podcast from the updated notebook?",
  "header": "Podcast",
  "options": [
    {"label": "Yes — deep dive", "description": "~10 min podcast covering all changes"},
    {"label": "Yes — brief", "description": "~3 min summary"},
    {"label": "No", "description": "Skip podcast generation"}
  ]
}])

if podcast_requested:
  mcp__notebooklm-mcp__studio_create(
    notebook_id=config.notebooklm_notebook_id,
    artifact_type="audio",
    audio_format=selected_format,
    confirm=True
  )

Output

Report what was synced:

Release Sync Complete — v{version}
  NotebookLM: source added to {notebook_title}
  HQ KB: ingested as release-notes/{version}
  Podcast: generating (poll with studio_status)