GEO / AI Search Toolkit

This directory is one public front door for the toolkit. Treat everything else in this package as internal support material, not as separate public skills.

What this skill owns

• site-specific GEO / AI-search audits
• single-page or section-level citability analysis
• AI crawler access, robots, renderability, and indexability checks
• llms.txt validation or generation
• structured data and entity-graph review
• platform-specific GEO notes for Google AI Overviews, ChatGPT web, Perplexity, Gemini, and Bing Copilot
• synthesis into remediation plans, executive summaries, JSON data, or PDF reports
• manual business-op flows only when the user explicitly asks for them

What this skill does not own by default

• generic SEO or marketing advice with no concrete site or artifact
• generic sales CRM work disconnected from a GEO engagement
• unsourced market-size claims, ROI claims, pricing claims, or invented benchmarks
• hidden slash-command routing to dozens of public sub-skills

Activation rules

Use this skill when at least one of these is true:

• the user provides a URL or domain and asks about GEO, AI search visibility, AI crawler access, schema, llms.txt, citability, or AI answer readiness
• the user wants a GEO remediation plan or client-ready report derived from actual audit findings
• the user explicitly asks for a GEO proposal, prospect update, or monthly delta comparison tied to a real site or prior audit

Do not activate this skill for:

• broad SEO discussion with no target site
• pure copywriting with no audit or implementation angle
• routine CRM bookkeeping unless the user explicitly asks for the GEO business workflow

Routing rules

Load the smallest set of worker files that can satisfy the task. Do not load the entire bundle by default.

• workers/audit.md for full or multi-domain audits
• workers/crawlers.md for robots, bot access, renderability, and indexability
• workers/technical.md for deeper rendering, meta, performance-risk, and technical analysis
• workers/citability.md for passage-level answer readiness
• workers/content.md for E-E-A-T, coverage depth, and content structure
• workers/schema.md for structured data and entity graph work
• workers/llmstxt.md for llms.txt validation or generation
• workers/brand-entity.md for brand/entity presence checks
• workers/platforms.md for platform-specific GEO notes
• workers/reporting.md for quick audit, full report, JSON, or PDF outputs
• workers/business-ops.md only when the user explicitly asks for prospect, proposal, or monthly comparison work

Use supporting references only when needed:

• references/workflow.md for audit mode selection and orchestration
• references/scoring-model.md for composite scoring and synthesis
• references/output-contracts.md for output shapes and filenames
• references/source-hygiene.md for claim discipline and evidence rules
• rubrics/*.md for scoring details
• templates/*.md and templates/*.json for output generation
• schemas/*.json for schema starting points
• scripts/*.py for deterministic extraction, scoring, llms.txt generation, and PDF rendering

Operating modes

1. Quick triage

Use when the user wants a fast read or a narrow answer.

1. Fetch the target with scripts/fetch_page.py or equivalent web retrieval.
2. Run only the relevant worker(s).
3. Output a compact summary with the top blockers, likely impact, and next actions.

2. Focused check

Use for one bounded area such as schema, llms.txt, bots, or citability.

1. Load the single worker plus its rubric if needed.
2. Avoid composite scoring unless the user explicitly wants it.
3. Return concrete fixes, not generic GEO theory.

3. Full audit

Use when the user wants a complete GEO assessment.

1. Start with workers/audit.md and references/workflow.md.
2. Run discovery on the homepage and any priority pages.
3. Pull in only the worker set justified by the site and request.
4. Synthesize with references/scoring-model.md.
5. Produce the output defined in references/output-contracts.md.

4. Reporting

Use workers/reporting.md plus the report templates when the user wants a client-facing or stakeholder-facing deliverable. Use scripts/generate_pdf_report.py only after the report JSON exists.

5. Manual business workflow

Only after explicit user intent, load workers/business-ops.md. These flows are manual and side-effectful by nature. Do not enter them implicitly.

Execution discipline

• Prefer deterministic scripts when they reduce extraction or formatting errors.
• Keep claims tied to observed site evidence unless external facts are freshly verified.
• Treat scripts/brand_scanner.py and scripts/citability_scorer.py as heuristics, not ground truth.
• Do not invent platform metrics, market stats, or ROI multipliers.
• Do not behave like a command index. This skill routes internally.

Optional platform adapters

• adapters/claude/ contains optional Claude-only subagents and install scripts.
• These adapters are not the source of truth. The canonical logic stays in this bundle.