Configure and Manage Feature Flags

This skill provides expertise in feature flag configuration, design patterns, and rollout strategies, with deep knowledge of feature flag systems like LaunchDarkly, Split, Unleash, ConfigCat, and custom implementations, covering flag lifecycles, targeting rules, gradual rollouts, and operational best practices. Core principles cover flag naming conventions (descriptive hierarchical naming like team.feature.variant, environment context where needed, avoiding negatives in flag names, versioning major iterations) and flag types (Kill Switch for emergency disable, Release Toggle for new feature visibility, Experiment Toggle for A/B testing, Ops Toggle for operational controls, Permission Toggle for access control).

Configuration structure is demonstrated through a basic YAML flag schema specifying description, type, default value, per-environment values, targeting rules by user segment or user ID, a percentage-based rollout with a ramp-up schedule, and metadata (owner, ticket reference, expiration date, tags). Advanced targeting rules are demonstrated through a JSON structure combining multiple conditions per rule (platform equals iOS and app version semver-greater-than-or-equal, or country in a list and user tier not free) each mapped to a variation and rollout percentage, with a default rule for unmatched users.

Implementation patterns cover multi-variate flags (a TypeScript interface for a checkout experiment with layout/paymentMethods/showTrustBadges variations retrieved via a typed getVariation call), dependency management (YAML requires/conflicts/suggests relationships between flags), and environment-specific configuration (different default-enabled and override behavior for development, staging mirroring production with exceptions, and production requiring approval with a max rollout rate per deployment).

Rollout strategies cover gradual rollout configuration (a staged percentage schedule with duration and success criteria like error rate and latency thresholds per stage, plus rollback triggers on error rate or user complaint thresholds within a time window) and ring-based deployment (canary ring for internal/beta users with auto-advance, early adopters ring for opt-in beta users, and general availability requiring manual approval). Monitoring and observability cover flag analytics configuration (tracking flag evaluations, user exposure, and performance metrics; custom events for feature engagement and conversion funnel tracking; dashboards with alert conditions on error rate and adoption rate thresholds).

Best practices cover flag lifecycle management (always including expiration date, owner, and success criteria at creation; monitoring metrics and documenting learnings while active; cleaning up code references and documenting outcomes at retirement; maintaining an audit trail of changes and approvals), security and compliance (separate namespaces for sensitive features, approval workflows for production changes, audit logging, data residency considerations, encrypting sensitive flag values), performance optimization (caching flag evaluations with appropriate TTL, local evaluation for high-frequency checks, circuit breakers for flag service failures, batching evaluations, monitoring evaluation latency against SLA thresholds), and team collaboration (shared naming conventions, common flag pattern libraries, flag templates, automated cleanup processes, documented flag dependencies).

Use this skill when configuring feature flags - naming and typing flags, writing targeting rules, designing a gradual or ring-based rollout, or setting up flag analytics and monitoring. It is well suited to teams using a feature flag system (LaunchDarkly, Split, Unleash, ConfigCat, or custom) for controlled feature releases. It is not meant for simple binary on/off toggles with no targeting, rollout, or lifecycle management needs.

Input: the feature being flagged, its target user segments, and the desired rollout strategy.

Output: feature flag configuration with naming, targeting rules, rollout schedule, and monitoring setup. Example gradual rollout stage:

- percentage: 5
  duration: "48h"
  successCriteria:
    errorRate: "<0.05%"
    conversionRate: ">baseline"

Works with feature flag platforms including LaunchDarkly, Split, Unleash, and ConfigCat, and integrates rollback triggers with monitoring/alerting systems like Slack notifications.

DevOps and product engineering teams configuring feature flags for controlled releases, and teams that need structured targeting, gradual rollout, and lifecycle management rather than ad hoc boolean toggles.