REQUIRED when adding OpenCode Zen models - provides a script that fetches accurate context, max tokens, pricing, and capabilities from the API. Use this skill whenever the user asks to add, register, or configure a new OpenCode Zen model. Do NOT guess model specifications—always fetch real data from the API.
REQUIRED when adding OpenRouter models - provides a script that fetches accurate context, max tokens, pricing, and capabilities from the API. Use this skill whenever the user asks to add, register, or configure a new OpenRouter model. Do NOT guess model specifications—always fetch real data from the API.
Debugging tools and techniques for acai. Use when investigating errors, analyzing session behavior, or troubleshooting issues in the acai CLI application. Includes guidance on reading logs, session files, and common error patterns.
Run fallow static analysis to find dead code, duplication, complexity issues, and architecture drift. Use before committing, after generating code, or when asked to clean up the codebase.
Survey a codebase as a senior advisor, find the highest-value improvement opportunities, present them for acceptance, and create ahm tasks for accepted findings. Strictly read-only on source code — never implements, fixes, or refactors anything itself.
Go through Open, Pending, and Blocked tasks in the backlog and make sure all decisions are made, dependencies are correct, and no open questions remain. The goal is to get tasks into a ready state for work.
Manually test acai features by running the app in tmux, interacting with the REPL, and inspecting logs and session files. Use after implementing a feature to verify it works end-to-end.
Migrate applications from AI SDK 6.x to AI SDK 7.0. Use when upgrading Vercel AI SDK packages, fixing v7 migration errors, or when the user mentions AI SDK v6, v7, upgrade, migration, breaking changes, system to instructions, fullStream, telemetry, tool context, or finalStep.
Run a focused review-readiness pass on a nearly finished change before commit. Scales the review to change size (XS/S = one pass, M = two passes, L/XL = three sequential passes covering rules conformance, correctness/source-of-truth, and overengineering). Then synthesize and apply the worthwhile fixes.