Zavelinski

Use when starting any non-trivial task to pick the BEST available capability (skill, plugin, MCP connector, agent, workflow) instead of the first one that works. Maps relevant skills/MCPs/agents, compares, picks the highest-quality fit, announces it, then executes. Trigger with /best-tool or whenever about to start a task by hand without checking for a better tool.

Use when the user says "godmode", "god mode", "/godmode", or wants max-capability mode for the session. Turns on exhaustive, best-tool-first behavior: map and use the BEST skill/plugin/MCP/agent, prefer Workflow/subagents for substantial work, verify before claiming done, and keep changes small and reversible. A UserPromptSubmit hook keeps it active every turn until "godmode off".

Use when the user says "godmode", "god mode", "/godmode", "ativa godmode", or wants max-capability mode for the session. Turns on exhaustive, best-tool-first behavior: map and use the BEST skill/plugin/MCP/agent, prefer Workflow/subagents for substantial work, verify before claiming done, persist context at the end. A UserPromptSubmit hook keeps it active every turn until "godmode off".

Use when the user says "salve" (or /salve) to ship and persist the current work in one shot. Persists context, commits, pushes to GitHub, opens and merges a PR into the main branch, triggers/confirms the deploy if the project has one, fast-forward-updates the local checkout, and reports evidence. Saying "salve" is the user's explicit OK to merge to the main branch and deploy. Variants - "salve doc" (doc-only, uses [skip ci], no redeploy) and "salve sem deploy" / "salve no deploy" (push + PR, do NOT merge to production).

Use quando o usuario disser "salve" (ou /salve) para entregar e persistir o trabalho atual de uma vez. Persiste contexto, commita, sobe no GitHub, abre e mergeia um PR na branch principal, dispara/confirma o deploy se o projeto tiver, atualiza a checkout local com fast-forward e reporta evidencia. Dizer "salve" e o OK explicito do usuario para mergear na branch principal e deployar. Variantes - "salve doc" (so-doc, usa [skip ci], sem redeploy) e "salve sem deploy" (push + PR, NAO mergear em producao).

When a task needs a specific format or style and zero-shot keeps drifting, give 2-5 carefully chosen examples instead of more instructions. Use for extraction/classification/formatting/tone tasks the model keeps getting subtly wrong. Trigger with /few-shot-curator or "add examples", "show it a few examples", "few-shot this".

To evaluate or compare outputs (candidate answers, PRs, drafts, model results) objectively, define an explicit rubric first and score each criterion with evidence, instead of a vibe "looks good". Use to pick the best of N options, gate quality, or grade generated content. Trigger with /rubric-judge or "score this", "judge against a rubric", "which option is best".

When two components/services talk (API client+server, producer+consumer, module boundary), pin the contract they share and test BOTH sides against it, instead of only unit-testing each in isolation. Use on integration boundaries, API changes, microservices, public interfaces. Trigger with /contract-test or "contract test", "will this break the consumer", "test the API boundary".

When you MUST feed a large blob into context (a long log, transcript, doc, or dataset dump), compress it to the salient parts first instead of pasting it whole. Use before including big inputs you cannot avoid. Distinct from context-warden (session discipline); this compresses one specific oversized input. Trigger with /prompt-compression or "compress this before adding", "summarize this blob for context", "this log is huge".

When an answer will be consumed by code or another step (not a human), define a schema first and make the output conform to it, instead of free text you then regex. Use for tool arguments, API payloads, extraction/classification results, agent-to-agent handoffs. Trigger with /structured-output or "make this JSON", "constrain to a schema", "structured result".

Before a commit, push, paste, or share, scan the diff/text for leaked secrets (API keys, tokens, private keys, passwords, connection strings) and block them, instead of leaking and rotating later. Use before any git push, before sharing logs/config, before pasting output. Trigger with /secret-scan or "check for secrets", "is there a key in this", "scan before push".

For a contested or high-stakes answer, have several agents propose answers then critique each other across a couple of rounds until they converge, instead of trusting one model or a blind vote. Use on hard design calls, ambiguous root-cause debates, or claims where reasoning quality matters. Trigger with /multi-agent-debate or "debate this", "have agents argue it out", "stress-test this answer".

For a complex problem, first decompose it into an ordered list of simpler subproblems, then solve them in sequence, feeding each answer into the next, instead of one-shotting the whole thing. Use on multi-step tasks, hard algorithmic problems, or anything that fails when attempted whole. Trigger with /least-to-most or "break this down", "decompose first", "solve step by step in order".

Before shipping a factual or analytical answer, draft it, generate independent verification questions about its own claims, answer those, and revise the draft to fix what the checks expose. Use on research answers, multi-claim summaries, audits, anything where a confident wrong detail is costly. Trigger with /chain-of-verification or "verify your own answer", "CoVe", "double-check each claim".

For factual or external claims, attach a source before asserting; mark anything you cannot verify as unverified; refuse to state ungrounded facts as if certain. Use when answering with facts, versions, APIs, prices, dates, or quotes - anywhere a confident-but-wrong claim is costly. Trigger with /cite-guard or "cite your sources", "is that grounded", "don't hallucinate this".