Cursor rule
Phase 1 - restate the task, surface unknowns, define testable success criteria, and classify complexity before any code.
3 2mo ago A 26 tokens
original MIT
Cursor rule
Phase 2 - gather real context. Read before editing, verify every API/symbol/path exists, never guess signatures.
3 2mo ago A 24 tokens
original MIT
Cursor rule
Phase 3 - design the approach with at least one alternative, state trade-offs, and enumerate edge cases and error paths before coding.
3 2mo ago A 27 tokens
original MIT
04-implement
04Cursor rule
Phase 4 - implement the chosen design as a minimal, incremental diff with no scope creep and no fake completions.
3 2mo ago A 24 tokens
original MIT
05-red-team
05Cursor rule
Phase 5 - adversarial self-critique. Assume bugs exist and actively try to break your own implementation before running it.
3 2mo ago A 26 tokens
original MIT
06-verify
06Cursor rule
Phase 6 - hard verification gate. Run typecheck, lint, and tests; paste real output; never claim success without evidence.
3 2mo ago A 27 tokens
original MIT
Cursor rule
Phase 7 - bounded retry loop. Diagnose root cause before re-editing; re-plan or escalate at iteration 3; stop and surface a blocker at iteration 5.
3 2mo ago A 35 tokens
original MIT
Cursor rule
When and how to spawn a readonly critic subagent for a fresh perspective, and how to fold its findings back into the pipeline.
3 2mo ago A 26 tokens
original MIT
orchestrator
09Cursor rule
Entry point for the disciplined reasoning pipeline. Invoke to make a cheap model plan, implement, self-critique, and verify with hard no-premature-done gates.
3 2mo ago A 34 tokens
original MIT
Cursor rule
Project-specific adapter — inject your commands, docs, and security gates into the generic pipeline phases.
3 2mo ago C 19 tokens
original MIT
Cursor rule
Cross-cutting protocol - offload durable state to a per-task memory file to virtually expand the context window and prevent decision drift.
3 2mo ago A 26 tokens
original MIT