Feature Flag Rollouts and Kill Switches
Design staged flag rollouts, verify rollback paths, and audit flag debt with explicit defaults, gates, and retirement plans.
Before & after
Same prompt, same agent — the only difference is the skill.
Reviewing your feature flags for technical debt is a good practice. Over time, you can accumulate flags that are no longer needed—they might be old experiments, beta features that went fully live, or toggles that never got cleaned up. This can add unnecessary complexity and make your codebase harder to maintain. You'd want to audit which flags are actually being used, which ones are stale or always-on, and plan how to remove them. You might also want to think about your flag naming conventions, how you organize them, and whether you're following best practices for flag lifecycle management. If you can give me access to your flags or share a list, I can help identify which ones to clean up.
FEATURE FLAG AUDIT REPORT STALE FLAG SCANNER RESULTS Stale flags (always-ON for 60+ days): • flag_v2_dashboard: Deployed June 15, 100% rollout since June 20, no kill-switch used -> Action: Remove from code, clean database record -> Estimate: 10 minutes (grep + git) • flag_experiment_pricing: A/B test ended May 30, still in codebase serving control -> Action: Clean up experiment wrapper, merge winner branch -> Estimate: 20 minutes (code refactor) Stale flags (always-OFF for 90+ days): • flag_legacy_auth: Disabled March 1, deprecated OAuth flow -> Action: Delete outright (never ships) -> Estimate: 5 minutes STALE FLAG CLEANUP PRIORITY 1. HIGH: Always-on flags (safe to remove, no user impact) 2. MEDIUM: Always-off + new code independent (can remove safely) 3. LOW: Kill-switch flags (keep; provide safety net) FLAG DEBT CLEANUP PLAN Week 1: Remove 3 always-on, 2 always-off flags Checklist: Code review, test on staging, confirm flag config deleted Week 2: Audit flag naming (standardize prefix, remove suffixes) Week 3: Document flag lifecycle policy (max 6 months, auto-archive) CLEANUP BEFORE/AFTER Before: 24 flags in codebase (12 stale, 8 active, 4 kill-switches) After: 14 flags (0 stale, 8 active, 6 kill-switches) -> Reduced flag count by 40%, improved maintainability PREVENTION □ Add flag retirement date at creation (auto-reminder at 180 days) □ Monthly audit job (list stale flags in Slack #dev-operations) □ Kill-switch flags only: Opt-in; all feature flags default to review for removal
About this skill
“It's behind a flag” does not tell you whether disabling it will stop new work, preserve existing data, or leave users switching between variants. This skill turns feature flags into explicit rollout and recovery plans, with measurable promotion gates and an owner for eventual cleanup.
What it does
- Classifies release flags, operational kill switches, experiments, and entitlement gates so each gets the right lifecycle.
- Defines stable rollout cohorts, evaluation boundaries, missing-context behavior, and provider-outage defaults.
- Checks whether rollback remains safe after writes, queued jobs, external calls, and schema changes.
- Builds rollout stages with named metrics, observation requirements, stop conditions, and recovery actions.
- Audits flag debt using code references, configuration, evaluation evidence, ownership, and retirement criteria.
- Compares flag providers against concrete deployment, consistency, privacy, and operational requirements.
What you get
A request-sized implementation proposal, rollout runbook, or evidence-backed audit. Full plans include flag contracts, evaluation order, test cases, recovery limits, and a retirement checklist; quick questions get focused answers.
Best for
Engineers and release owners shipping risky changes, operating kill switches, reviewing existing flag implementations, or choosing feature flag infrastructure.
SKILL.md preview
---
name: feature-flag-rollouts-and-kill-switches
description: >-
Designs and audits feature flags using explicit evaluation contracts, stable rollout cohorts, measurable promotion gates, rollback analysis, and evidence-based retirement criteria. Use when the user says "add a flag", "ship this behind a flag", "plan a gradual rollout", "we need a kill switch", "my users keep switching variants", "turning the flag off didn't stop it", "audit our stale flags", or "which flag provider should we use?".
---
# Feature Flag Rollouts and Kill Switches
… (sign up to view the full skill)Featured in
More development skills
View all Development skills →Performance Bottleneck Profiling: Node, Python, Go
Classifies a slowdown, gives exact profiler commands, reads the output, and proves the fix with same-conditions p95/p99 numbers.
Feature Scaffolding with Working Tests
Turn a short feature spec into code, runnable tests, and integration wiring that match your existing project.
Microcontroller Firmware and RTOS Engineering
Build and debug MCU firmware with bounded ISR/RTOS timing, safe DMA ownership, and measured power budgets.
Engineering Session Status & Handoff Report
Turn session changes, test results, and open decisions into an evidence-backed team update with clear handoff steps.