From supervibe
Use BEFORE shipping a risky feature TO design staged rollout (kill-switch / percentage / cohort), define rollback criteria, plan flag debt cleanup. Triggers: 'feature flag', 'staged rollout', 'постепенный релиз', 'фича-флаг'.
npx claudepluginhub vtrka/supervibe --plugin supervibeThis skill is limited to using the following tools:
BEFORE merging the first commit that gates new behaviour behind a flag. BEFORE expanding an existing flag to a wider audience. WHEN an incident requires emergency disable. WHEN flag debt is suspected (flag count rising, cleanup PRs absent).
Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
BEFORE merging the first commit that gates new behaviour behind a flag. BEFORE expanding an existing flag to a wider audience. WHEN an incident requires emergency disable. WHEN flag debt is suspected (flag count rising, cleanup PRs absent).
This skill picks the rollout shape (kill-switch, percentage, cohort), defines the rollback contract, and locks in a cleanup deadline. Without all three, a flag is a long-lived branch in production and accrues debt.
Follow docs/references/skill-expert-operating-standard.md: start from source of truth, preserve retrieval evidence, apply scope safety, use real producers with runtime receipts for durable delegated outputs, verify before completion claims, and keep confidence below gate when evidence is partial.
grep -r "isEnabled\|flag(\|getVariation\|featureFlag" src/. Flag count + median age tells you whether debt is already a problem.What is the worst plausible outcome of this flag flipping ON?
Data corruption / billing error / security regression → KILL-SWITCH only, server-evaluated, default OFF
User-visible bug for some users → PERCENTAGE rollout 1% → 5% → 25% → 50% → 100%
Different value for different audiences (plan, geo) → COHORT targeting with explicit segment definition
Unknown — instrumenting a hypothesis → DARK LAUNCH (run new code, ignore output, compare)
Is rollback automated?
YES → require a synthetic alarm bound to the flag's success metric
NO → require a named on-call owner and a tested manual disable path
Is this a long-lived config switch (not a release gate)?
YES → it is NOT a feature flag; move it to typed config + PRD decision section
NO → continue, with cleanup deadline
supervibe:confidence-scoring with artifact-type=agent-output; ≥9 required to mark this skill complete.Flag: <name> Owner: <person> Cleanup deadline: <YYYY-MM-DD>
Shape: <kill-switch | percentage | cohort | dark-launch>
SLI: <metric + window + threshold>
Stages: 1% (24h) → 5% (48h) → 25% (48h) → 50% (24h) → 100%
Advance gate: <inequality on SLI>
Rollback gate: <inequality on SLI> OR <qualitative tripwire>
Kill-switch: tested-in-staging=<date>; runbook=<link>
Cohorts (if any): <segment name → membership rule>
Cleanup PR: <draft link or tracking issue>
Storage backend: <platform / repo path> (NOT env vars)
supervibe:test-strategy — defines the contract test pinned to the OFF branch.supervibe:incident-response — consumes the kill-switch + runbook designed here.supervibe:prd — capture the rollout decision as an architecture decision record.supervibe:audit — periodic flag-debt sweep.