Cursor rule
Agents are not humans — do not estimate timelines, effort, or team size using human baselines.
AI-native development methodology — rules, skills, and scripts for working with coding agents
Cursor rule
Agents are not humans — do not estimate timelines, effort, or team size using human baselines.
Cursor rule
How agents pick up, verify, and close beads — JIT verification, self-review against ACs, evidence policy, and bd remember knowledge capture.
Cursor rule
Standards for creating and closing beads — every bead must be self-contained with ACs, and every close must include evidence.
Cursor rule
Convention for marking deliberate simplifications with ceiling and upgrade trigger. Prevents "later means never" by requiring each deferral to name its limitation and the condition for revisiting.
Cursor rule
When and how to create committed design specs for larger initiatives.
Cursor rule
Multi-agent code review protocol — same-model multi-lens (autonomous) + cross-model (human-assisted) for structural changes.
Cursor rule
Planner/Executor operating model — role definitions, workflow, scratchpad conventions, and beads task tracking.
Cursor rule
Prevents file-edit conflicts between the parent agent and running subagents — do not touch files in a subagent's blast radius until it completes.
Cursor rule
Lazy senior dev implementation mode — the decision ladder for minimal, correct code. Activate with 'use ponytail-playbook' or when implementing feature/task beads where code minimalism matters. Defers to pragmatic-tdd.mdc for test discipline.
Cursor rule
Signal-first TDD — test discipline varies by bead type. Prevents zero-signal tests and over-testing.
Cursor rule
Session start/close protocol — the first and last things an agent does in every session. Mandatory.
Cursor rule
Git worktree isolation, shared beads DB, and worktree lifecycle for multi-branch development.