Executing SDDs

Assign an SDD

/sdd-assign SDD-001

You'll be asked to choose an agent backend:

/sdd-assign updates the SDD frontmatter (agent, assigned_to, status: assigned) and launches the appropriate workflow.

Interactive execution (Claude Code)

/sdd-assign SDD-001 claude

Claude loads the SDD into the current session and starts implementing. You can guide, review, and adjust in real time. The agent fills the Work Log when done.

Autonomous execution (OpenHands)

/sdd-assign SDD-001 openhands

OpenHands must be running first:

mise run openhands:up

Then trigger execution:

mise run sdd .forgia/sdd/FD-001/SDD-001.md

OpenHands runs headlessly in Docker, reads the SDD, implements the component, and fills the Work Log.

Batch execution

Run all pending SDDs in an FD sequentially:

forgia batch FD-001

Forgia runs each pending SDD in order, one at a time. It stops immediately on the first failure — fix the failing SDD and re-run before continuing with the rest.

Direct execution

forgia exec .forgia/sdd/FD-001/SDD-001.md

Runs a single SDD through the configured agent runtime without the /sdd-assign flow.

Work Log

Every agent must fill the Work Log before marking the SDD done:

• Executor: who or what ran the SDD
• Decisions: choices made during implementation (with reasons)
• Output: files created, commands run, test results
• Retrospective: what worked, what didn't, what the next agent should know

An SDD with an empty Work Log fails /fd-verify.