pocket-cfo

Use for active docs, CFO Wiki docs, local-dev docs, eval docs, benchmark docs, roadmap, START_HERE, or docs that guide future Codex threads.

Use for proof bundles, evidence summaries, alert cards, monitoring results, reporting outputs, approval cards, validation ladders, or shipped-claim truthfulness.

Use when a task is complex, spans multiple files or packages, or needs a step-by-step execution document. Create or update a Finance Plan in plans/ that follows PLANS.md exactly, uses the F0-F6 roadmap, and remains live during implementation.

Use for F6 planning or implementation involving monitors, alert cards, thresholds, investigation missions, policy/control ownership, recurring checks, or demo replays.

Use for any Pocket CFO task that creates, updates, audits, or executes Finance Plans, active roadmap docs, phase handoffs, or plan truthfulness. Do not use to widen scope beyond the current active plan.

Use when implementing GitHub webhooks, repository sync, installation auth, or repo-content ingestion. Enforces GitHub App-first integration, least privilege, webhook signature verification, installation-token caching, and connector-only scope inside Pocket CFO.

Use for implementation or planning tasks that touch module boundaries, route design, schema, runtime, approvals, reporting, monitoring, or package-level architecture.

Use at the end of every Pocket CFO Codex slice or when reviewing Codex output. Produces a compact handoff packet for human review and next-prompt design.

Use for tasks touching source ingest, finance twin facts, discovery answers, reporting artifacts, monitoring, proof bundles, freshness, lineage, or CFO Wiki truth.

Use whenever a Pocket CFO prompt asks for validation, smoke ladders, QA reruns, or reproducibility gates.