specification: API Commons Agent Skills
specificationVersion: '0.1'
provider: LaunchDarkly
providerId: launchdarkly
generated: '2026-08-27'
method: generated
source: >-
  Authored by the API Evangelist enrichment pipeline. Every operationId referenced in
  every skill was verified present in openapi/launchdarkly-rest-api-openapi.json,
  harvested 2026-08-27 from https://app.launchdarkly.com/api/v2/openapi.json (HTTP 200,
  OpenAPI 3.0.3, 401 operations). No operation, header, error code or reversal window in
  these skills is invented; each traces to that contract, to
  conventions/launchdarkly-conventions.yml, or to a docs URL cited in the skill.
note: >-
  LaunchDarkly publishes no AGENTS.md or first-party skill pack of its own. It does
  publish a 125-tool hosted MCP server, which is the provider's own answer to this
  problem — see mcp/launchdarkly-mcp.yml. These skills cover the REST path, which is what
  an agent uses when it is not going through MCP, when it needs an operation the MCP
  server does not expose (348 of 401 operations have no tool), or when it is running on
  the federal or EU instance where the hosted MCP server is unavailable.
skills:
  - name: launchdarkly-ship-a-feature-flag
    file: launchdarkly-ship-a-feature-flag.md
    description: >-
      Create a flag, target it in one environment, turn it on, and confirm. The core
      CodeControl loop, with the semantic-patch Content-Type, the no-Bearer-prefix
      authorization rule and the 30-day reversal window all stated inline.
    operations: [getProjects, getProject, getEnvironment, postFeatureFlag, getFeatureFlag, patchFeatureFlag, getFeatureFlagStatus]
    risk: write
  - name: launchdarkly-governed-flag-change
    file: launchdarkly-governed-flag-change.md
    description: >-
      Propose a change through an approval request or a scheduled change instead of
      applying it directly. The safe write path for an agent in production, and the only
      one with a clean withdrawal step.
    operations: [postApprovalRequest, getApprovalRequests, getApprovalRequest, postApprovalRequestApply, deleteApprovalRequest, postFlagConfigScheduledChanges, getFlagConfigScheduledChanges, deleteFlagConfigScheduledChanges, getAuditLogEntries]
    risk: write
    entitlement: Enterprise
  - name: launchdarkly-audit-and-clean-up-flags
    file: launchdarkly-audit-and-clean-up-flags.md
    description: >-
      Find stale flags, check evaluation status in every environment and check dependent
      flags before retiring anything, then deprecate or archive rather than delete.
    operations: [getFeatureFlags, getFeatureFlag, getFeatureFlagStatuses, getFeatureFlagStatus, getDependentFlags, patchFeatureFlag, deleteFeatureFlag, getAuditLogEntries]
    risk: read-then-write
cross_references:
  conventions: conventions/launchdarkly-conventions.yml
  errors: errors/launchdarkly-problem-types.yml
  rate_limits: rate-limits/launchdarkly-rate-limits.yml
  authentication: authentication/launchdarkly-authentication.yml
  mcp: mcp/launchdarkly-mcp.yml