Explore possibilities for a problem or opportunity with structured evaluation. Do NOT enter plan mode — this skill produces a brainstorm artifact directly. Triggers: /brainstorm, brainstorm, explore options, what are our options.
Review implementation code against the plan for drift, gaps, and blind spots. Triggers: /code-review, review code, check implementation, compare to plan, code vs plan.
Record, look up, or reconcile entries in the decision ledger — the persistent log of decided truths. Triggers: /decide, record decision, log this decision, what did we decide, decision ledger, supersede decision.
Create a technical architecture and design document. Do NOT enter plan mode — this skill produces a design artifact directly. Triggers: /design, design this, architecture for, technical design.
Create or expand a structured implementation plan with phases, tasks, subtasks, and verification criteria. Re-running on an existing plan deepens it — adds tasks, fills gaps, refines subtasks. Do NOT enter plan mode — this skill produces plan artifacts directly. Triggers: /plan, create a plan, plan this…
Adversarial critical analysis of plans, specs, or designs. Triggers: /poke-holes, poke holes, find gaps, challenge this, what could go wrong, devil's advocate.
Write a requirements specification for a feature. Do NOT enter plan mode — this skill produces a spec artifact directly. Triggers: /specify, write spec, specify requirements, requirements for.
Structural verification and quality patterns for C and C++ — sanitizers (ASAN/UBSAN/TSAN/MSAN), const correctness, ownership, thread safety, modern idioms. Load when planning, implementing, or reviewing C/C++ code (.c, .cc, .cpp, .h, .hpp; CMake, Makefile).
Recording user decisions as durable truth in the decision ledger. Load whenever the user makes a design or architecture choice, defines a project concept or term, answers a design question, reverses an earlier decision, or when current work touches a topic the ledger may already govern — including plain conversation…
Structural verification for Rust — clippy with warnings-as-errors, miri for unsafe/FFI. Load when planning, implementing, or reviewing Rust code (.rs; Cargo.toml).
Structural verification for Swift — SwiftLint, TSAN for concurrency. Load when planning, implementing, or reviewing Swift code (.swift; Package.swift, .xcodeproj).