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.mjs validates every skill: frontmatter, required sections, EN/ES template parity, voice rules, and that Related skills links resolve.
node scripts/build-codex.mjs writes the flat Codex layout to dist/codex/. Add --check --codex-repo=<path> to fail on drift against a checkout of the Codex repo.
node scripts/eval.mjs runs the LLM judge over evals/<skill>/prompts.jsonl using your signed-in claude CLI session (no API key). Set EVAL_CLI=codex to 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. Use DRY_RUN=1 to 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.