Project Memory Workflow

What Problem It Solves

After long sessions, context resets, or AI-to-AI handoffs, projects can become a pile of scattered chat with no clear target, completed work, blockers, or next action. daily-workflow writes project state into a local Docs/ folder so the next session or next AI can resume directly.

How It Works

The skill maintains a set of workflow files around start, checkpoint, wrap-up, and handoff triggers:

• PROJECT.md stores project metadata, phase, and last checkpoint.
• TARGET.md holds purpose, success criteria, and scope; it is read-only by default after creation.
• STATUS.md records current state, compressed context, risks, and recent updates.
• COMPLETED.md appends completed items without overwriting history.
• PENDING.md tracks unresolved work and blockers.
• NEXT_ACTIONS.md is rewritten at each checkpoint or wrap-up as the current action plan.

The workflow compresses the conversation into intent, decisions, command or test outcomes, file paths, blockers, and next steps instead of pasting the full transcript. Legacy files are read first, migrated or merged with confirmation, and never auto-deleted. Older status entries can be archived to Docs/archive/YYYY-MM.md.

Boundaries

It fits engineering or operations projects that need resumable state across sessions, especially multi-person or multi-AI handoffs. It is not a place for secrets, customer privacy, full confidential documents, or large logs. If the target is unconfirmed, the skill preserves that uncertainty instead of inventing results.