Project Plans
A DAG of Features, with one task DAG inside each Feature.
Keep the hierarchy simple
Project Plan → Feature → Task. SQLite holds the authoritative state, dependencies, journal entries, and execution history. You can start with one Feature and one Task and grow the plan as the work requires.
Plan once, continue by ID
/spoq-plan "Build checkout with an API and web interface"
# Continue with the saved Project Plan ID
/spoq checkoutThe IDs shown are illustrative; continue with the IDs returned by your planner. The planner saves the smallest useful graph and runs planning validation. Root SPOQ continues with the same IDs, claims ready Tasks, invokes implementation validation, and records the result. In Codex, use $spoq-plan and $spoq skill mentions.
Inspect both levels
Alternative: from your project root, build this example manually in an empty store instead of invoking the planner above.
spoq create project_plan checkout "Checkout"
spoq create feature api "API" --parent checkout
spoq create feature web "Web interface" --parent checkout
spoq depend web api
spoq create task contract "API contract and tests" --parent api
spoq dag checkout
spoq dag apiFeature dependencies also gate Tasks in downstream Features. Static waves describe ordering; readiness reflects current statuses, prerequisites, and claims. Reopened upstream work blocks dependent dispatch.
Work DAGs and strategy graphs answer different questions
Project Plan and Feature DAGs describe what depends on what. A strategy graph describes how planning, review and implementation proceed for that saved scope. A PIQARD rejection revisits a planning phase within a bounded loop; it does not create a cycle in your task prerequisites. See both views in the workflow guide.
Coming from maps and epics?
Legacy maps convert to Project Plans and epics convert to Features. Tasks retain their dependencies. Conversion requires confirmation and keeps original files plus byte-for-byte backups. Review the upgrade guide.