AgentKit
DocsKitsCLI ReferenceDesktop App
BetaYou are reading docs for the beta channel (2.13.0-beta.20). Features may change before the next stable release.Switch to stable →

Skills

Produce fair competitor intelligence with ak:competitor

Research a bounded competitor set, preserve source provenance, separate facts from inferences, and create reviewable analyses, battlecards, or comparison-page drafts.

Use ak:competitor to investigate competitor positioning, content, search presence, and comparison opportunities. The useful outcome is a dated, source-backed decision artifact, not an unsupported winner or a promise that a comparison page will rank or convert.

Choose the action that matches the decision

Use ak:competitor when

  • analyze [url] should examine positioning, pricing, audience, channels, strengths, and weaknesses.
  • content [url] should identify content topics, formats, and gaps.
  • seo [url] should compare available keyword, backlink, authority, and content evidence.
  • alternatives [competitor] should draft a singular alternative, roundup, you-versus-competitor, or competitor-versus-competitor page.
  • list should show competitors already represented in the working context.

Choose another workflow when

  • You need a wider audience or market synthesis rather than one competitor set. Use ak:marketing-research.
  • You need to define a customer profile from customer evidence. Use ak:persona.
  • You already have approved intelligence and need a campaign decision. Use ak:marketing-planning.

Prepare evidence boundaries

Provide your product facts, the competitor names or URLs, geography, audience, comparison criteria, and the date range that counts as current. Identify which sources may be used and whether the runtime may access the network.

Do not use customer quotes, pricing, rankings, feature availability, reviews, or migration claims without a source and observation date. Mark each conclusion as a sourced fact, an inference from facts, your supplied assumption, or an unresolved question.

Invoke a bounded action

RuntimeInvocationAvailability boundary
Claude Code/ak:competitor <action> ...Can use available web and specialist-agent capabilities; access is not guaranteed.
Cursor/ak:competitor <action> ...Slash invocation is user-verified; tools and provider behavior can differ.
Codex$ak:competitor <action> ...Uses native Skill discovery; network and projected-agent availability remain session-specific.
/ak:competitor analyze https://example-competitor.com "Compare positioning, public pricing, audience, and migration claims for US small teams. Cite current primary sources, include observation dates, and label every inference"

The canonical actions are analyze, content, seo, alternatives, and list. Use the exact action name so the intended branch is clear.

Understand the stages

  1. Set the comparison frame. Confirm product, competitors, market, audience, timeframe, decision, and permitted sources.
  2. Gather current evidence. The workflow can use web search and specialist Agents for research, positioning, search analysis, or battlecards when those capabilities are available.
  3. Cross-check claims. Prefer first-party product, pricing, policy, and changelog sources; use independent reviews as attributed perspective, not product truth.
  4. Analyze the selected action. Compare like-for-like criteria, state missing data, and keep inference separate from observation.
  5. Create the decision artifact. The base workflow writes an analysis under reports/competitors/{date}-{name}.md; battlecards may use assets/sales/battlecards/{competitor}.md. The alternatives route can also create centralized competitor data and page drafts organized by format.
  6. Review before downstream use. Recheck time-sensitive claims and obtain legal, brand, sales, SEO, and publication approval as applicable.

Keep research and external action separate

Comparative content must stay accurate and fair

A competitor's public statement is evidence of what it states, not proof of universal performance. A generated strength, weakness, score, testimonial, or recommendation is not publishable until verified and reviewed.

Network research reads external services and may expose queries or URLs to the selected provider. Do not bypass access controls, submit private material, or collect personal data without authority. This Skill does not authorize ad spend, competitor-keyword bidding, outreach, account changes, tracking pixels, page publication, or migration offers.

Verify outputs and evidence

A useful report or page draft should include:

  • Scope, audience, criteria, observation date, search terms, and sources.
  • A fact table that links each material claim to current evidence.
  • Inferences, assumptions, conflicts, and unknowns shown separately.
  • Fair “best for” and “not ideal for” judgments for every compared option.
  • Decision implications and next checks, not guaranteed rankings, revenue, or conversion.
  • The exact created artifact paths and any publication or refresh approval still outstanding.

Troubleshoot and know the limits

SymptomSafe next step
Web or Agent tools are unavailableAsk for approved source material or produce a research plan with explicit evidence gaps.
Pricing or feature claims conflictPrefer dated first-party sources, cite both versions, and leave the conflict unresolved until verified.
SEO metrics are missingDo not invent rankings, backlinks, search volume, or domain authority; name the unavailable dataset or provider.
alternatives lacks your product factsSupply approved positioning, pricing, ideal use cases, limits, and migration evidence before drafting copy.
A draft is ready to publishRun brand, legal, factual, SEO, accessibility, and owner review in a separate approved step.

The Skill does not define a live competitor database, guaranteed data provider, refresh service, or measurement methodology. Comparison-page structures are templates, and their keyword or conversion outcomes are unproven. Continue with the Marketing Kit overview or Runtime adapters.