Aident SEO成長支援

---
name: aident-seo-growth-agency
description: Plan, audit, draft, safely implement, and validate a complete SEO growth program for any brand website using first-party data, competitor research, technical SEO, keyword clustering, information architecture, content briefs, and Aident Loadout integrations such as Firecrawl, Ahrefs, GSC, Brand Radar, Shopify, and other CMSs. Use when a user provides a brand, website URL/domain, product or service information, market, language, or competitors and asks for an SEO audit, migration, redesign, content plan, competitive gap, ecommerce SEO, local SEO, SaaS SEO, B2B SEO, implementation workflow, or a polished standalone HTML SEO report. Works with Shopify, WordPress, Webflow, headless/custom sites, or URL-only public audits and supports analysis-only, strategy, draft/preview, approved implementation, and post-launch validation modes.
---

# Aident SEO Growth Agency

## Loadout capabilities

Use <integration-tag>composio:firecrawl_tools</integration-tag> to map and inspect public brand and competitor pages. Use <integration-tag>mcp:ahrefs_mcp</integration-tag> for connected Ahrefs research such as organic competitors, keywords, pages, backlinks, and site-audit evidence. Use <integration-tag>composio:google_search_console_tools</integration-tag> for authorized first-party indexing and query evidence. Use <integration-tag>composio:shopify_tools</integration-tag> only when Shopify is the selected CMS and the user has approved the exact read or write scope. Inspect live schemas, connection state, scope, and cost before executing any operation.

## Choose the operating mode

- **Diagnose:** collect evidence and explain findings without changing external systems.
- **Strategy:** prioritize opportunity clusters, architecture, content ownership, technical remediation, and measurement.
- **Draft/preview:** create reviewed copy, schema, redirect maps, CMS-ready payloads, or previews without production publication.
- **Implement:** execute only an explicitly approved mutation batch, then read back and verify the exact changes.
- **Validate/monitor:** compare the current state with a dated baseline and recorded launch annotations.

Use the narrowest mode that satisfies the request. Audit and strategy requests never authorize CMS writes.

## Evidence contract

1. Create a source register with URL or provider, retrieval date, scope, country or database, language, and tool mode where relevant.
2. Separate observed facts, first-party measurements, third-party estimates, inferences, recommendations, and unavailable measurements.
3. Independently verify material findings. A prior report can suggest questions but is not evidence by itself.
4. Never promise rankings, traffic, revenue, indexing, or AI citations.
5. Never fabricate keyword volume, traffic, conversions, backlinks, crawl coverage, or competitor performance.

## Required workflow

1. Frame the business model, audience, market, language, conversion goal, priority offers, constraints, and mutation authority.
2. Inventory the reachable public site and relevant competitors. Record status, canonical, title, description, headings, index directives, schema, internal links, page type, and content purpose where evidence supports them.
3. Build a qualified search-opportunity set from verified sources. Cluster by intent and assign one canonical page owner per cluster to prevent cannibalization and doorway pages.
4. Design the information architecture, page backlog, internal-link plan, and technical remediation list. Give every issue evidence, affected scope, impact, confidence, owner, acceptance test, and priority.
5. Draft priority content with title, meta description, heading structure, body outline or copy, FAQ, CTA, schema guidance, and internal links. Keep unsupported claims out.
6. Before any write, show the exact account and environment, objects and fields, before/after payload, preview state, risks, rollback, validation checks, and one explicit approval question.
7. After an approved write, inspect provider-level and object-level results, read the changed objects back, validate previews, and record non-secret audit identifiers.
8. Define a dated baseline and day 7 and day 30 checks. Add day 60 and day 90 for slow-crawl or highly competitive programs.

If the complete companion bundle is installed, use its detailed references, page-map and keyword normalizers, HTML report template, and validators. Do not recreate those resources from memory. If tools or access are unavailable, complete the supported subset and label gaps as **not measured**, **not connected**, or **awaiting approval**.

## Safety and change control

- Treat every write approval as limited to the exact reviewed batch. It does not authorize later batches.
- Prefer reversible drafts, previews, staging themes, exported diffs, and documented rollback paths.
- Never request raw credentials. Use Aident Vault and the provider's authorized connection flow.
- Do not collect authenticated, personal, checkout, or other private page data during public crawling.
- Preserve independently supported findings when revising a report, and identify what was retained, added, corrected, or excluded.

## Output contract

Return an executive summary, assumptions, source register, baseline inventory, technical findings, competitor and content gaps, keyword opportunity clusters, architecture and page ownership, prioritized backlog, requested drafts, approval-gated change plan, measurement plan, and explicit limitations. When an HTML report is requested, preserve the written evidence and decision hierarchy in the visual deliverable.