From mainbranch
Provides a deterministic Main Branch status briefing inside Claude Code. Runs `mb status --json --peek` and summarizes what changed, what is healthy, what is stale, and ranked next actions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/mainbranch:mb-statusThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Show the operator the deterministic Main Branch briefing without duplicating
Show the operator the deterministic Main Branch briefing without duplicating repo-health checks in prose.
CLI facts first: Run mb status --json --peek before answering, then use
that JSON as the source of truth.
Shared source: The portable daily start/status workflow contract lives in
workflows/mb-start-status/workflow.md. This Claude skill is the Claude Code
status shell over that source.
Shared contract markers: Keep these aligned with the shared source.
Required commands:
mb status --json --peekmb start --jsonmb doctor repair --planmb doctor repair --plan is read-only. If it exits nonzero while returning plan_interpretation and actions, treat that as findings to review, not as an opaque command failure.
Required fact paths:
money_pathmoney_path.policymoney_path.policy.thresholds_declaredmoney_path.active_betsmoney_path.active_bets.unanchoredmoney_path.active_bets.over_capmoney_path.objects.proof.qualityvalidation.file_contractscontent_strategyranked_actionsupdatereadinessreadiness.dimensions.repo_runtimereadiness.dimensions.business_memorydrift.itemsruntime.codex_cliruntime.claude_codesince_last_checkjournalcheckpointonboardingintegrationsgithubtopology.repo_boundarybrain.betsbrain.bets.activebrain.bets.due_soonbrain.bets.overduebrain.bets.exit_criteriabrain.bets.exit_criteria.missingbrain.bets.exit_criteria.triggered_failure_signalsbrain.bets.exit_criteria.triggered_double_down_signalsvocabularyApproval gates: updates_repairs_migrations, file_writes, checkpoint,
provider_mutation, publishing_or_spend, customer_contact, private_data,
destructive_operations, and status_marker.
Public/private boundaries: no_secrets, no_raw_provider_exports,
no_customer_member_data, no_private_runtime_settings, and
no_raw_finance_legal_records.
Core route: status facts first, runtime mismatch gates before business routing, one owner-facing recommendation, business language first, and explicit approval before mutating the status marker or doing durable writes.
cd into
the business repo or pass the repo path.mb status --json --peek
readiness.dimensions.repo_runtime,
readiness.dimensions.business_memory, vocabulary, content_strategy,
money_path, validation.file_contracts, and ranked actions.ranked_actions first. For each one, include:
money_path when the question is about the path from customer progress
to offer, proof, CTA, channel, push, playbook, page readiness, or outcome
feedback. Keep the language evidence-based: legible, supported, connected,
instrumented. Do not say the offer is good, bad, likely to convert, or ready
to win.
Use validation.file_contracts when the question is about a file that
exists but lacks the business shape needed for the next workflow. Route
offer-shape gaps to /mb-think and ask before editing durable offer files.
Use content_strategy when the question is about content strategy health,
layered channel/account/person files, stale platform rules, or disconnected
content layers. Do not infer that health by parsing markdown yourself unless
status says the section is unavailable.
Use brain.bets.active, brain.bets.due_soon, brain.bets.overdue, and
brain.bets.exit_criteria for active, due-soon, overdue, missing-exit,
triggered kill, triggered double-down, close, update, and narrate moments
before inventing bet state from prose.
Use topology.repo_boundary when the operator is deciding whether work
belongs in this business repo, a separate business repo, or a
child repo.
If vocabulary.terms.push defines display words, use them in
operator-facing prose without changing current paths, frontmatter, JSON
keys, validator rules, or command names.
Use team facts and GitHub activity author_display / author_known fields
when naming people; keep unknown contributors as handles.
Do not re-run shell probes that duplicate status facts.
For "what changed?" or "what happened since last time?", answer from
since_last_check.journal first, then top-level journal for recent
context.integrations.github and
integrations.providers from status first. If the operator needs choices or
repair commands, run:mb connect plan
mb connect doctor --json
Summarize GitHub, Cloudflare, Google/Workspace, Meta Ads, and Apify as numbered
business choices. Use the CLI's next_command or repair_command; do not ask
the operator to paste tokens into files or public issue text.
--peek is the default inside this skill because a conversation may inspect
status before the operator is ready to record a daily check.
If the operator explicitly says this is the daily check-in and wants it recorded, run:
mb status --json
Use the new report for the answer.
Respect each action and signal's safe_to_share field. If safe_to_share is
false, keep evidence local to the conversation and do not suggest pasting it
into public GitHub without review.
npx claudepluginhub noontide-co/mainbranch --plugin mainbranchDetects repository state and user intent to route to appropriate skills or CLI commands for setup, sync, repair, or launch.
Shows a single-screen work status dashboard with task state, epic progress, knowledge flywheel health, and recent activity. Useful for getting oriented or deciding next actions.
Generates CTO-level project status from git log/status and key docs (README, TODO, CLAUDE.md). Summarizes shipped work, progress, blocks, next steps for standups or 'where are we' queries.