AI AGENT
SKILLS

Create clean SKILL.md folders with focused instructions, references, examples, and optional scripts.

Use when: You want to turn a repeatable workflow into a portable agent skill.

Review third-party skills for suspicious instructions, scripts, secrets, network calls, and unsafe scope.

Use when: You are about to install or publish a third-party SKILL.md package.

Work with PDFs, spreadsheets, Word docs, and presentations using agent-readable office workflows.

Use when: You need agents to inspect, transform, or produce office files.

Run a local web app, inspect browser output, capture screenshots, and debug UI behavior.

Use when: You need an agent to verify that a web app actually renders and works.

Inspect failing GitHub Actions logs, identify the likely cause, patch the issue, and verify the fix.

Use when: A pull request or main branch has failing GitHub checks.

Collect actionable PR review comments, implement fixes, and report what changed.

Use when: You need an agent to handle review feedback without losing thread context.

Review React and Next.js code for rendering, accessibility, performance, and maintainability.

Use when: You are building or reviewing modern React/Next.js interfaces.

Code review, TDD, debugging, migrations, simplification, and engineering documentation workflows.

Use when: You want a broad but practical software-agent operating system.

Deploy, inspect, optimize, and debug Vercel-hosted apps with agent-readable workflows.

Use when: You need an agent to work with Vercel projects or deployment output.

Work with models, datasets, Spaces, training jobs, evaluation, and Hugging Face Hub workflows.

Use when: You need agent help around model or dataset engineering.

Plan and run Playwright, Cypress, Selenium, Appium, API, and CI testing workflows.

Use when: You need cross-browser, mobile, API, or CI testing support.

Review SQL for grain, joins, filters, nulls, date logic, dialect issues, and performance risks.

Use when: A query is going into a report, dashboard, model, or stakeholder analysis.

Turn messy metric debates into canonical definitions with grain, owner, caveats, and examples.

Use when: Different teams are using the same metric name to mean different things.

Check datasets for missingness, duplicates, stale dates, broken joins, and suspicious distributions.

Use when: You need to trust a dataset before using it for a decision.

Review dashboards for metric accuracy, filters, stale data, misleading charts, and broken interactions.

Use when: A dashboard is going to executives, clients, or an operating team.

Analyze A/B tests with hypotheses, guardrails, exposure checks, segments, and decision notes.

Use when: You need a clean readout from an experiment without overreading noisy slices.

Triage data freshness, volume, schema, lineage, and quality incidents.

Use when: A pipeline or dashboard looks wrong and you need a source-backed incident summary.

Use agent skills for search, observability, Elastic workflows, and data infrastructure tasks.

Use when: Your stack uses Elastic for search, logs, or observability.

Analyze public companies, SEC filings, business segments, estimates, ratings, and risk factors.

Use when: You need source-grounded public-company research.

Review MD&A, footnotes, risk factors, accounting policies, and filing changes.

Use when: You need an agent to summarize or compare public filing sections.

Draft an investment memo with thesis, market, traction, risks, diligence gaps, and recommendation.

Use when: You need a first-pass IC memo from notes, decks, filings, and diligence material.

Audit assumptions, formulas, links, balance checks, sensitivities, and model documentation.

Use when: A spreadsheet model is about to be used for a real decision.

Use agent skills for trading research, risk review, portfolio notes, and market-analysis workflows.

Use when: You are using an agent for trading research or portfolio operations, not financial advice.

Review commercial contracts against a playbook, flag issues, and suggest fallback language.

Use when: You need a structured first-pass contract review before counsel review.

Review an NDA for obligations, scope, duration, exceptions, remedies, and escalation items.

Use when: Business teams need an NDA triage before legal signoff.

Extract clauses, obligations, dates, risks, and citations from many contracts into a table.

Use when: You need diligence-style clause extraction with source locations.

Review data-processing workflows, DPAs, privacy obligations, DSARs, and compliance gaps.

Use when: A product, vendor, or process touches personal data.

Adversarially review AI-generated legal work for citations, unsupported claims, and risk.

Use when: An agent produced legal analysis, contract notes, or a compliance memo.

Search medical literature, retrieve citations, and organize evidence for research workflows.

Use when: You need a transparent literature-search process with citation discipline.

Draft clinical or research protocols with objectives, methods, outcomes, risks, and reporting needs.

Use when: You need a structured protocol draft from research notes and source material.

Check research writeups against reporting standards such as PRISMA and EQUATOR-style guidance.

Use when: A manuscript, protocol, review, or evidence summary needs reporting-quality checks.

Use a broad healthcare and biomedical SKILL.md catalog selectively for FDA, trials, guidelines, and drug workflows.

Use when: You need healthcare research workflow coverage beyond a single protocol or literature review.

Turn clinician-approved source material into plain-language patient education drafts.

Use when: You need understandable patient-facing copy without giving diagnosis or treatment advice.

Use focused marketing agent skills for product marketing, SEO, AEO, CRO, copy, launch, and analytics.

Use when: You need a broad marketing workflow pack with practical operator coverage.

Create campaign briefs, landing page audits, AEO checks, ad variants, and voice-of-customer summaries.

Use when: You need agent workflows for GTM execution, not just content drafts.

Run growth experiments, content operations, outbound systems, SEO workflows, and revenue-intelligence tasks.

Use when: You want a more operational growth and marketing automation skill pack.

Review pages for search visibility, AI answer readiness, structured data, entities, and content gaps.

Use when: A page needs to be discoverable in search and AI-assisted answers.

Research an account, company, or buyer before outreach, meetings, or deal review.

Use when: A seller, founder, or CS lead needs a fast account brief.

Prepare agendas, discovery questions, context, risks, and likely objections before a customer call.

Use when: You have CRM notes, transcripts, email context, or prior support history to synthesize.

Turn transcripts or notes into decisions, risks, next steps, owner/date actions, and CRM-ready summaries.

Use when: You need follow-ups after sales, customer success, or discovery calls.

Classify, prioritize, route, deduplicate, and draft next actions for support tickets.

Use when: A support queue needs fast, consistent triage.

Package customer impact, timeline, reproduction steps, failed workarounds, and needed decision.

Use when: Support needs to escalate an issue to product, engineering, or leadership.

Turn resolved tickets, workarounds, and internal notes into customer-facing help content.

Use when: The same support issue is repeating and should become reusable documentation.

Clean duplicate, stale, inconsistent, and poorly routed CRM records.

Use when: The CRM is becoming a decorative database nobody trusts.

Write product requirements with problem, users, scope, non-goals, metrics, risks, and rollout plan.

Use when: A feature idea needs to become a clear buildable product brief.

Turn interview transcripts into themes, evidence, confidence, quotes, and product implications.

Use when: You have user research notes and need decisions, not a transcript smoothie.

Compare roadmap bets using frameworks, assumptions, evidence, dependencies, and decision notes.

Use when: A product team needs to choose what not to build.

Create an experiment plan with hypothesis, assignment, primary metric, guardrails, and readout plan.

Use when: A team wants to test a product change without kidding itself.

Review UI screens for clarity, hierarchy, spacing, accessibility, consistency, and likely usability issues.

Use when: You need a structured design pass before shipping UI.

Review interfaces against usability heuristics and accessibility basics with severity labels.

Use when: A product needs a UX audit grounded in established heuristics.

Turn messy process notes into an SOP with owner, trigger, steps, exceptions, controls, and revision history.

Use when: A process is repeated often enough that improvising it is expensive.

Compare documented process against actual evidence, bottlenecks, handoffs, controls, and failures.

Use when: The process doc says one thing and the team is clearly doing seventeen others.

Create role-specific competencies, evidence anchors, rating scales, and debrief rules.

Use when: A hiring team needs structured evaluation instead of vibes with a calendar invite.

Build interviewer guides with agenda, competency questions, follow-ups, rating anchors, and feedback forms.

Use when: Multiple interviewers need to evaluate the same role consistently.

Prepare meeting context, decisions needed, likely questions, risks, and owner/date action logs.

Use when: A meeting needs to become useful before everyone loses another hour of life.

Compare vendors using weighted criteria, must-haves, TCO, security risk, references, and recommendation.

Use when: A team is choosing a tool and needs evidence instead of demo-induced hypnosis.

Rewrite HR or admin policies in plain English while preserving obligations, exceptions, and risk flags.

Use when: A policy is technically written in English but spiritually a compliance cave.