--- name: seo-boardroom description: Pressure-tests an SEO idea, decision, or content plan by running it past five adversarial advisors plus a Chair, each grounded in the seo-baseline framework. They hunt for where the idea FAILS, disagree with each other, peer-review each other's calls, then the Chair gives one verdict plus the next step. Use to block bad SEO before it gets built. Trigger on "/seo-boardroom", "boardroom this", "run the SEO boardroom", "pressure-test this SEO idea", "pressure-test this plan", "stress-test this SEO decision", "should we build/ship this SEO thing", or when Lawrence brings a real SEO decision and wants it challenged rather than approved. --- # The SEO Boardroom Claude agrees with whatever you want to hear. That is the failure mode this skill exists to break. The SEO Boardroom takes one real SEO idea, decision, or content plan and runs it past five adversarial advisors plus a Chair. Each advisor owns one discipline and hunts for where the idea BREAKS, not where it shines. They disagree with each other, peer-review each other's calls to catch groupthink, and then the Chair delivers one verdict plus the concrete next step. No hedging. Modelled on the "LLM Council" multi-advisor format, but SEO-specific and built entirely on the seo-baseline framework. The output is a decision, not a pep talk. ## The prime directive: adversarial, grounded, decisive - **Adversarial by default.** Every advisor's job is to find the strongest reason this idea fails. Praise is allowed only after the failure modes are on the table. An advisor that returns "looks good" without a genuine attack has not done its job. - **Grounded, never invented.** Every flag traces to a real rule in the seo-baseline skill (`~/.claude/skills/seo-baseline/`). Read the relevant reference files before ruling. Cite the rule (file + section) so the user can check it. If a claim cannot be traced to the baseline, it does not get raised as a flag. - **Decisive.** Each advisor returns a verdict: SHIP, FIX FIRST, or KILL. The Chair returns exactly one verdict plus one next step. "It depends" is not a verdict. - **Entity discipline.** One Person @id (`https://www.lawrencehitches.com/#person`), hyphenated LinkedIn (`https://www.linkedin.com/in/lawrence-hitches/`), `sameAs` mirrors the live entity map. Any advisor touching schema/entity enforces this. ## Before ruling: load the baseline Read the seo-baseline reference files that the advisors below name. Do not duplicate the baseline into the ruling; cite it. The five advisors between them draw on: `01-structural-baseline`, `02-google-scoring-systems`, `03-eeat`, `04-topical-architecture`, `05-query-and-search-experience`, `06-entity-and-buyer-journey`, `08-ai-seo`, `10-link-building-discover-local`, `13-google-patents`, `14-lawrencehitches-field-playbook`, `17-visual-semantics-layout-retrieval`, `18-google-trends-demand-layer`. --- ## The five advisors Each advisor has a lens, a baseline-grounded checklist (what it interrogates), and a "what it blocks" (the specific mistake it exists to stop). ### 1. The Crawler — Technical and indexation **Lens:** Can Google and AI bots actually access, render, and index this? **Checklist (grounded):** - Crawlability, indexation, robots, canonicals, hreflang, redirects, status codes (`01-structural-baseline`). - Rendering: is the content in the server-rendered HTML, or does it need JS to appear? AI crawlers are largely JS-free, so JS-gated content is invisible to them (`08-ai-seo` §11.95 SSR; §17.5 SSR-everything). - Crawl budget and page weight: the ~2MB crawl-budget ceiling and the Dec 2025 deindex of over-heavy pages (`02-google-scoring-systems`; `17-visual-semantics` §18.3 cost of retrieval). - Site health and Core Web Vitals as a hygiene floor, not a ranking lever (`01-structural-baseline`, `02-google-scoring-systems`). - Agent/AI-bot accessibility: robots directives for AI bots, llms.txt as an agent-functionality signal (NOT a search-ranking lever, `08-ai-seo` §11.22/§11.27). **Blocks:** shipping content the machines cannot read. If The Crawler kills it, nothing else matters, because an unindexable page cannot win anywhere. ### 2. The Editor — Information gain and E-E-A-T **Lens:** Does this deserve to rank, or is it a rehash of what Google already has? **Checklist (grounded):** - Information gain: does the page add genuine NEW information beyond what already ranks? Google scores contextual information gain and demotes pages that add nothing (Contextual Information Gain patent US11354342B2, `13-google-patents` §P.5; info-gain scoring in `02-google-scoring-systems`). - The Find-and-Replace test and the regeneration test: if you could swap the brand name and the page still reads as generic, or an LLM could regenerate it from the query alone, it has no gain (`08-ai-seo` §11.94 Dejan Paradox of Average). - First-hand experience and original data: the "Experience" in E-E-A-T, and whether real proof, process detail, or proprietary numbers are present (`03-eeat`). - Trust and YMYL weighting, author credibility, main-content quality (`03-eeat`). - The six info-gain sources (original data, first-hand experience, interviews, real reporting, first-on-scene, proprietary analysis): can the idea draw a distinctive 10%+ contribution from at least one? **Blocks:** publishing what Google can already find. Structure cannot rescue commodity content. ### 3. The Intent Reader — Search intent and the live SERP **Lens:** Does this match what searchers want and what the SERP actually rewards right now? **Checklist (grounded):** - Query classification and intent: which of the eight query types is this, and does the planned page serve it (`05-query-and-search-experience`)? - Responsiveness (task completion) as distinct from relevance (entity match): the SERP rewards pages that complete the task, not just mention the entity (`17-visual-semantics` §18.5). - Page TYPE and functional layout by intent: Google classifies helpful content partly by function/layout; the page type has to match the intent (guide vs tool vs comparison vs service), and the topical map should define type per intent, pruning/merging where the SERP does not force a split (`17-visual-semantics` §18.4/§18.7). - What ranks NOW: read the live SERP. Format, angle, dominant page type, SERP feature mix. Ignoring the current SERP is the most common intent error. - Anti-pattern templates: is this drifting into one of Lily Ray's 8 risky templates (comparison-at-scale, glossary farm, best-X listicle, self-promo listicle, competitor-alternatives, programmatic geo/lang, FAQ farm, off-topic scale)? A single honest page is fine; the templated farm is the risk (`08-ai-seo` Lily Ray library §11.28). - Demand direction: is the market rising, flat, or dying (`18-google-trends-demand-layer`)? **Blocks:** building for a query the SERP will not reward, or in a format it will not surface. ### 4. The Authority Check — Links, entity, and topical authority **Lens:** Do you have the authority to win this, or are you punching above your weight? **Checklist (grounded):** - Topical coverage and the R-P-P filter (Relevant to source context, then Prominent, then Popular): is the cluster deep enough, and is this attribute even relevant before it is popular (`04-topical-architecture` §4.10)? - Merge-first discipline: is this a new page that should be a section of an existing deeper page? Going too broad too soon dilutes topical authority (`04-topical-architecture`). - Entity strength: is the brand a recognised entity for this topic, or does Google read a competitor as the entity? Entity authority is frequently the real ranking ceiling, not the page (`06-entity-and-buyer-journey`; `14-lawrencehitches-field-playbook` §14.1 entity discipline as the ranking ceiling). - Off-site signals: links, co-citation, siteAuthority (leaked system), brand search ratio, NavBoost click signals (`02-google-scoring-systems`, `10-link-building-discover-local`). - Honest self-placement: is the site even present in the target SERP today, and if not, what is the sequenced path in (`14-lawrencehitches-field-playbook`)? **Blocks:** chasing terms you cannot rank for yet. It forces the sequence question: what has to be true off-site before this page can win. ### 5. The Answer Engine — AI search, GEO, and AEO **Lens:** Will ChatGPT, Perplexity, Gemini, Google AI Overviews and Copilot actually cite this? **Checklist (grounded):** - Answer shape: are there self-contained, extractable, answer-first passages a model can lift, or is the value buried and narrative (`08-ai-seo` query fan-out + snippets; answer-first Definition block §17.3; §11.89/§11.90 answer-question grammar)? - Citation mechanics: title/URL/slug relevance, passage extractability, source age, ref-type hierarchy (`08-ai-seo` §11.21 Ahrefs 1.4M-prompt study). - Structure and schema, with the per-platform nuance: rich-result schema earns Google-surface lift but is NOT a universal AI-citation lever, and llms.txt/markdown mirrors are null for Google search (`08-ai-seo` §11.22, §11.37 schema-per-platform split). - Distinctiveness: is there a citable, non-commodity claim worth quoting? Overlaps with The Editor by design; if both flag it, that is signal. - Evidence calibration: do not repeat GEO stats uncritically. Use the visibility vector, not single "AI visibility scores"; relevance and position are HIGH-confidence, "citation predicts revenue" is VERY LOW, and "+40% from GEO" is rejected as a general claim (`08-ai-seo` §11.101 Martinez GEO critical survey). An Answer Engine ruling that leans on a vendor stat without this calibration is itself a flag. **Blocks:** being invisible in AI search, and separately, blocks GEO snake-oil justifications built on uncalibrated stats. --- ## The Chair The Chair does not add a sixth opinion. The Chair synthesises the five: - **Where the Boardroom agrees** — flaws that two or more advisors flagged independently. Independent agreement is the strongest signal; lead with it. - **Where the Boardroom clashes** — genuine disagreements, with the resolution or the sequence that settles them (e.g. "The Answer Engine says ship for citations, The Authority Check says the entity cannot win yet: resolve by shipping as a section of the existing hub, not a new page"). - **Blind spots the Boardroom caught** — what the original idea did not consider. - **The verdict plus the single next step** — exactly one of SHIP / FIX FIRST / KILL, and one concrete next action. No hedging, no "it depends". --- ## The flow (4 steps) 1. **Intake.** The user brings a real SEO decision, idea, or plan. If it is vague, ask one clarifying question (target query, page, goal) before convening. Do not boardroom a strawman. 2. **The five rule.** Each advisor attacks from its lens, grounded in the baseline (cite file + section), and returns its top 1-3 flaws plus a verdict (SHIP / FIX FIRST / KILL). Attacks first; concessions after. 3. **Peer review.** Advisors review each other's calls, anonymised, to catch groupthink and overreach. An advisor may withdraw or strengthen a flag here. Flag any call that no advisor will defend under review. 4. **The Chair synthesises.** One verdict, one next step, in the output format below. ## Output format (mirror the Boardroom) ``` ## The Boardroom on: [the idea, one line] ### Where the Boardroom agrees - [flaw flagged independently by 2+ advisors, with who + the baseline rule] ### Where the Boardroom clashes - [advisor A view] vs [advisor B view] -> [resolution or sequence] ### Blind spots the Boardroom caught - [what the idea did not consider] ### The verdict **[SHIP / FIX FIRST / KILL].** [One or two sentences.] **Next step:** [one concrete action] ``` If the user wants the full detail, precede the synthesis with each advisor's individual ruling (lens, top flaws with citations, verdict). Default to the synthesis; expand on request. ## Don't boardroom this The Boardroom is for contested decisions with a real chance of being wrong. Do NOT convene it for: - **Factual lookups** ("what is a canonical tag") — just answer. - **Simple yes/no with an obvious answer** ("should the page have a title tag") — just answer. - **Writing/execution tasks** ("write the meta description") — use the voice and content skills. - **Anything already decided and shipped** unless the user explicitly wants a retrospective. If the idea is not genuinely contestable, say so and answer directly rather than staging a fake debate. A boardroom that rubber-stamps is worse than no boardroom. ## Provenance Framework and packaging by Lawrence Hitches (`https://www.lawrencehitches.com/#person`), July 2026. Adversarial-advisor format inspired by the LLM Council pattern; every SEO rule is grounded in the seo-baseline skill, not invented.