go

name	go
classification	active
description	Re-ground and structure execution with ACTION/REVIEW/GROUNDING cadence. Use mid-session to realign with the current plan and priority order.

GO GO GO

A grounding workflow for AI agents working in this template repository. This workflow structures task execution and periodic self-assessment. It complements AGENT.md, which provides the canonical directives, rules, and architectural context.

Read ALL of this document, then carry out the Action.

Ground Yourself

Read .agent/directives/AGENT.md and follow all instructions. Read .agent/directives/principles.md and reflect on the rules.

Read .agent/memory/README.md, .agent/memory/active/distilled.md, .agent/memory/active/napkin.md, and .agent/memory/operational/repo-continuity.md (if they exist) to absorb context from prior sessions.

If your current task is driven by a plan, read the relevant domain current/README.md first, then the plan before executing the next step.

If you discover bugs, regressions, broken quality gates, or contradictory canon while grounding, those take priority over every other lane.

If no bugs are present and the entry surfaces still point at unfinished completion work, treat that unfinished work as the default priority before starting new feature or research work. Check the relevant domain current/README.md to see whether a live active tranche exists.

If the task is runtime infrastructure work, read .agent/plans/runtime-infrastructure/current/README.md before taking action.

If the task is demo application work, read .agent/plans/demo-application/current/README.md before taking action.

Intent

• Identify and state the current plan you are working to. What impact does the plan seek to bring about?
• State explicitly whether this session is:
  • bug fixing,
  • unfinished completion work, or
  • genuinely new work. Default to unfinished completion work when no bugs exist and the current entry surfaces still point there.
• What are you trying to achieve? Take a step back and consider the big picture, think hard about it, and then reflect on your thoughts. Has anything changed? Why?

Structure the Todo List

• Your todo list must achieve the intent of the plan. Populate it with tasks that are atomic, specific, measurable, provable, and ACTIONABLE. Make each task small enough for the result to be easily and comprehensively reviewed. All actions must be prefixed with ACTION:.
• Start by proving the priority order:
  • if bugs exist, the first ACTION: items must triage or fix them
  • if no bugs exist and unfinished completion work is live, the first ACTION: items must advance that completion plan
• If you have tasks that are large or complex, break them down into smaller, more manageable tasks.
• Immediately after each ACTION: there MUST be a REVIEW: item. This consists of:
  1. Stepping back and reflecting on the action
  2. Checking alignment with the plan and rules
  3. Checking whether the action changed continuity or active-memory surfaces
  4. Invoking the appropriate sub-agent(s) per the reviewer catalogue in executive memory
• Make sure your todo list includes running the quality gates. These items should be prefixed with QUALITY-GATE: and happen reasonably often.
• Periodically include a GROUNDING: task to re-read this document and .agent/directives/AGENT.md, ensuring your todo list stays relevant and aligned with the plan.
• Every fourth REVIEW: should be a holistic review invoking multiple sub-agents to assess overall coherence.
• Remove any items from your todo list that don't make sense, or are no longer relevant.
• Every 6th item must be to re-read .agent/skills/go/SKILL.md and re-follow ALL instructions in this file, including this one.

Action

Please start the next task in the todo list, and carry on.