harness-hook

Agent lifecycle hooks can be managed across Claude Code, Copilot, and Codex by defining a canonical source of truth in .harness/src/hooks/ that deterministically translates into each provider's specific configuration format.

Synchronizing hook configurations across multiple coding agents is best handled by storing canonical definitions in JSON files and using an apply command to project them into provider-specific outputs, ensuring consistency across environments.

To scaffold and deploy a lifecycle hook, run npx harness add hook to generate the JSON definition, edit the file, and execute npx harness apply to translate and deploy the configuration to all supported providers.

Yes, hook projection enforces structural validation by checking the canonical shape, events matrix, and handler types during the apply process to satisfy functional requirements before generating provider outputs.

Yes, canonical hook definitions stored in .harness/src/hooks/ are automatically translated into each provider's config format during the apply step, eliminating the need to manually edit individual provider configurations.

A single source of truth for agent lifecycle hooks is needed because managing configurations independently across multiple providers is complex, error-prone, and leads to inconsistencies between different development environments.