codes1gn

Enhance prompts for Claude models (Opus, Sonnet, Haiku). Applies XML structuring, recency-optimized instruction placement, few-shot patterns, and Claude-specific best practices.

Enhance prompts for OpenAI GPT models and Codex. Applies primacy-optimized instruction placement, conversational structure, delimiters, and GPT/Codex-specific best practices.

ACTION-FORCING prompt enhancer router. Detects /enhance-me → launches Task subagent → displays enhanced prompt → executes it. Reading this skill without launching a subagent is a protocol violation.

Durable request skill — keeps every AI agent conversation alive across multiple tasks instead of silently finishing. AGENT SHOULD STRICTLY FOLLOW THIS PROTOCOL WHEN ACTIVATED. ONLY activate when the user explicitly invokes.

Durable request skill — keeps every AI agent conversation alive across multiple tasks instead of silently finishing. AGENT SHOULD STRICTLY FOLLOW THIS PROTOCOL WHEN ACTIVATED. ONLY activate when the user explicitly invokes.

/deep-sleep — Keep agent alive while user is away. Enter low-power polling loop that prevents request timeout.

ACTION-FORCING prompt enhancer router. Detects /enhance-me → launches Task subagent → displays enhanced prompt → executes it. Reading this skill without launching a subagent is a protocol violation.

Systematic auditor and improver for agent skill suites that use harness scripts. Use when the user wants to improve reliability, reduce token waste, tighten determinism, or strengthen capability of any multi-skill system that orchestrates agent workflows via CLI scripts. Trigger on phrases like "boost harness", "improve skills", "audit harness", "skill reliability", "reduce token waste", "harness best practices", or "/boost-harness". This skill is explicitly triggered only — NEVER auto-activate during a tuning or workflow execution.

CroqTuner harness — robust FSM-driven GPU kernel tuning loop with guard flags, idea dedup, and compaction-safe state. Wraps ai-tune-from-current-best and ai-tune-from-scratch with mechanical enforcement of the tuning protocol. Use when the user asks to "tune", "ai-tune", "croqtune", "tune from best", "tune from scratch", or "start tuning".

Sweep ALL scenario shapes and adapt the reference best kernel (f16 iter143, e4m3 iter068) to each shape with up to 30 iterations per shape. Non-stop until every shape across all scenarios has a tuned variant. Use when the user asks to "tune from best", "adapt kernels", "sweep tune", or "ai-tune-from-current-best".

Tune typical shapes from scratch (baseline .co kernels) for >=150 iterations each. Covers shapes far from the reference that need full optimization. Non-stop sweep until all typical shapes are tuned. Use when the user asks to "tune from scratch", "full tune", or "ai-tune-from-scratch".

Save the current session state for later resumption. Captures session progress, decisions, blockers, next actions, and memory extractions into a structured XML file. Use when the user says "handoff", "save session", "wrap up", or "end session".

Resume a previously saved session. Loads context from a handoff XML file and renders a visual summary report so you can continue exactly where you left off. Use when the user says "resume", "continue session", "load session", or "resume [tag]".

/incubate — Fully autonomous project incubation pipeline. Takes an idea through PRD → Architecture → Development → Testing → README → Website with persistent cross-session memory.

Save the current session state for later resumption. Captures session progress, decisions, blockers, next actions, and memory extractions into a structured XML file. Use when the user says "handoff", "save session", "wrap up", or "end session".

Checkpoint-based agent workflow inspired by Feynman's explain-like-a-child technique. Forces every agent turn that completes work to present an interactive checkpoint using the environment's builtin interactive tool instead of silently finishing. Applies to ALL tasks — coding, writing, debugging, research, analysis, refactoring, any prompt. Use always, on every task completion, before ending any turn where work was performed.