Operator Skills para Claude
federicoimparatta/operator-skills
Método estruturado para product, delivery, analysis e ops
Ver no GitHubConjunto de skills para builders e operadores que implementam metodologia estruturada em especificação de produtos, escopo de entregas, análise técnica e fluxos operacionais. Cobre PRD, feature proposals, análise de dados, auditorias UX, roadmaps, OKRs, entrevistas, competitive teardown, scoping, migrações, dashboards, RFPs e avaliações.
Instalação
npx convez-ai federicoimparatta/operator-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Escrever PRDs como hipóteses testáveis com escopo e critérios de aceitação
- Scopear trabalho de delivery customizado com fases e modelos comerciais explícitos
- Sintetizar entrevistas em padrões e jobs, separando comportamento de preferência
- Fazer auditoria UX heurística e gerar tickets priorizados por severidade
- Definir roadmaps a partir de dashboards e comportamento de usuários
- Responder RFPs de governo e enterprise com matriz de conformidade estruturada
Veredito

Skills fortemente estruturadas e bilíngues (EN/ES) que operacionalizam método rigoroso de product e delivery. Sem templates genéricos ou conteúdo vazado; foca em regras, templates e filosofia — ideal para equipes que priorizam método sobre estilo e querem reduzir backlog theater.
README
A tradução para o português deste README ainda está na fila.
operator-skills
Claude skills for builders and operators. Product specification, delivery scoping, technical analysis, and operational workflows. Method over voice. Structure over style.
Built from the method behind Manual de ataraxia digital: imperturbabilidad entre tanto scrum by Federico Imparatta (Montevideo, 2025). The handbook provides the philosophy; these skills are its operational expression.
Principles
- Self-contained. Every skill is a standalone unit with its own voice rules and templates. No shared imports between skills.
- Bilingual. Every template ships in English and Rioplatense Spanish.
- Templates and rules only. No real examples, no leaked client content, no redacted artifacts.
- Method over voice. These skills work regardless of the operator's own tone.
- Philosophy-anchored. Every skill names the concepts it operationalizes.
Philosophy
See PHILOSOPHY.md for the ten core concepts that shape every skill in this repo.
Skills
product
- prd-authoring. Write a PRD as a hypothesis, not a spec. Problem, user, scope, acceptance criteria, rollout, instrumentation, open questions.
- feature-proposal. Pitch a feature before scoping. Problem framing, proposed direction, expected impact, cost to validate.
- analytics-instrumentation. Event taxonomy, naming conventions, funnel definitions, and instrumentation from scratch.
- ux-audit. Heuristic audit method that produces actionable tickets ranked by severity and surface area.
- roadmap-from-data. Convert dashboards and user behavior into a prioritized sequence. Avoid backlog theater.
- okr-drafting. Objectives as outcomes, key results as evidence. Name the bet, set kill criteria, reject task lists.
- interview-synthesis. Turn interview notes into patterns and jobs. Separate behavior from preference, kill the assumption you loved.
- competitive-teardown. Study what a competitor shipped, not what they claim. Feature matrix, what they bet on, the gap you can own.
delivery
- proposal-scoping. Scope custom dev work. Phases, commercial model, explicit in-scope and out-of-scope.
- migration-spec. Platform-to-platform migration method. Inventory, parity matrix, cutover, rollback.
- dashboard-spec. Specify internal tools and admin dashboards. Data model, views, roles, states, empty and error handling.
- rfp-response. Structured response to government and enterprise RFPs. Compliance matrix, technical approach, team, pricing.
- security-questionnaire. Answer vendor security assessments without overclaiming.
analysis
- data-reconciliation. When two sources disagree, find the gap. Schema, time window, definition, sampling, root cause.
- red-team-review. Adversarial review of proposals, architectures, agent systems. Structured failure modes, not vibes.
- candidate-evaluation. Score a candidate against a rubric of judgment, delivery, depth, ownership, failure fluency, communication.
- vendor-evaluation. Score a vendor on delivery risk, pricing fairness, stack fit, commercial terms, references, switching cost.
- post-mortem. Retrospective method for shipped work, failed work, or incidents. Timeline, contributing factors, action items, no blame.
- decision-record. Lightweight ADR for product and ops decisions. Options, the bet, consequences accepted, revisit trigger.
ops
- agenda-synthesis. Synthesize a daily or weekly agenda across calendars and orgs.
- ticket-triage. Linear and Jira triage. Categorization, sequencing, and when to decline.
- documentation-extraction. Turn closed tickets, threads, and conversations into functional docs.
- meeting-prep. Briefing before a meeting. Stakeholders, history, desired outcome, objections, opening move.
- sprint-planning. Capacity after meetings and on-call, sequence by uncertainty, commitment vs forecast, named owners.
comms
- status-update. Weekly or cycle status update. Lead with the outcome that moved, not the task list.
- exec-brief. One-page brief for leadership or the board. The bet, the number, the ask.
- stakeholder-update. Async delivery-status update to a client or sponsor, plus an escalation note for slips.
- incident-comms. Stakeholder-facing comms during an incident. Impact, status, next update time, no blame in real time.
Voice rules
Every skill in this repo applies the same voice rules. They are the through-line.
- No buzzwords: leverage, empower, synergy, resonate, tapestry, delve, elevate, captivate, explore, dynamic, testament.
- No em dashes.
- No filler openings.
- Decision-oriented over descriptive.
- High-context, no overexplaining.
- Standard sentence case for formal outputs.
- ES outputs match the register of the handbook: direct, self-aware, no reverence for process, comfortable naming failure.
- EN outputs use direct US business register. Do not soften bluntness into corporate-speak.
- Language switches based on audience, not author.
Installing
These skills are designed for Claude Code, but the method and templates are tool-agnostic.
Claude Code
Clone into your project's .claude/skills/ directory or reference individual skills directly:
git clone https://github.com/federicoimparatta/operator-skills.git ~/operator-skills
Then either:
- Per-project. Symlink or copy the skills you want into
.claude/skills/in a specific project. - Global. Symlink or copy into
~/.claude/skills/to make them available everywhere.
Each skill is a folder with a SKILL.md that Claude Code loads on demand.
Other tools
The templates in each templates/ folder are plain markdown. Use them with any assistant or none.
Maintenance
This repo is the source of truth. The Codex variant (operator-skills-codex) is generated, not hand-edited.
node scripts/lint.mjsvalidates every skill: frontmatter, required sections, EN/ES template parity, voice rules, and thatRelated skillslinks resolve.node scripts/build-codex.mjswrites the flat Codex layout todist/codex/. Add--check --codex-repo=<path>to fail on drift against a checkout of the Codex repo.node scripts/eval.mjsruns the LLM judge overevals/<skill>/prompts.jsonlusing your signed-inclaudeCLI session (no API key). SetEVAL_CLI=codexto use Codex instead; it skips cleanly if neither CLI is on PATH.scripts/push-codex.sh ["message"]releases the Codex variant in one step: lint, generate, clone the Codex repo, sync, verify zero drift, commit, and push. UseDRY_RUN=1to preview without pushing.
To release the Codex variant, run scripts/push-codex.sh. Do not edit the Codex repo by hand.
Contributing
See CONTRIBUTING.md.
Credits
Drawn from Manual de ataraxia digital: imperturbabilidad entre tanto scrum (Federico Imparatta, Montevideo, 2025). Over 10,000 copies sold in Spanish.
License
MIT. See LICENSE.