功能描述

Use when Ray asks about the OpenClaw SEO-GEO Workflow, ClawLite/OpenClaw daily SEO/GEO workflow, Hunter/Tony/Peter blog production, daily blog factory delive...

使用说明 (SKILL.md)

OpenClaw SEO-GEO Workflow

Name: OpenClaw SEO-GEO Workflow
Author: x-rayluan

This is the controlling runbook for Ray's OpenClaw SEO-GEO daily workflow. It coordinates Hunter discovery, Tony content production, the 12-post SEO/GEO factory, Peter publish/build/live QA, and daily SEO patrol for ClawLite/OpenClaw marketing.

Use this skill as the first stop. Do not invent a parallel workflow. The source of truth is the deterministic scripts and receipts under /Users/m1/.openclaw/workspace.

Default Stance

Receipt first, then claims. Inspect durable JSON/Markdown receipts before saying a workflow passed, failed, published, or used a skill.
Local staging is not production publishing. STAGED_LOCAL, Peter local build success, and source publish are not live deployment.
Production publish claims require deploy evidence plus live QA.
Monitor/ranking/backlink/AI-citation claims require connector or live patrol data.
Missing live dependencies should degrade to explicit SKIPPED, NEEDS_DATA, or PASS_WITH_WARNINGS, not silent success.
Never let the legacy /Users/m1/Projects/clawlite project take over clawlite.ai production aliases unless Ray explicitly asks.
Never write into /Users/m1/Projects/clawlite/content/blog from chat-triggered workflow runs unless Ray explicitly asks for main-site local apply and the command uses --allow-main-site-write.

Main Commands

Run from /Users/m1/.openclaw/workspace.

Full daily workflow, recommended default:

node scripts/run-hunter-tony-seo-geo-workflow.mjs --date YYYY-MM-DD

Factory-only workflow:

node scripts/run-daily-seo-geo-blog-factory.mjs --date YYYY-MM-DD

Safe full workflow with no live deploy:

node scripts/run-hunter-tony-seo-geo-workflow.mjs \
  --date YYYY-MM-DD \
  --peter-skip-live-deploy \
  --peter-skip-live-qa

Protected main-site local apply requires an explicit override. Do not add this flag unless Ray asked to modify the main ClawLite repo:

node scripts/run-clawlite-publish-apply.mjs \
  --date YYYY-MM-DD \
  --allow-main-site-write

Live deploy path requires explicit evidence:

node scripts/run-hunter-tony-seo-geo-workflow.mjs \
  --date YYYY-MM-DD \
  --peter-allow-live-deploy \
  --peter-deploy-command "npm exec -- vercel deploy --prod --prebuilt" \
  --peter-deploy-evidence-path /path/to/deploy-evidence.json \
  --peter-live-qa-base-url https://clawlite.ai

Environment equivalents:

CLAWLITE_PETER_DEPLOY_COMMAND
CLAWLITE_PETER_DEPLOY_EVIDENCE_PATH
CLAWLITE_PETER_SKIP_LIVE_DEPLOY=1
OPENCLAW_SEO_PATROL_MODE=live
DATAFORSEO_B64

Role Boundaries

Hunter: topic radar, multi-source discovery, keyword opportunity scouting, raw scouting receipts, durable wiki handoff briefs, and 14-30 day dedupe.
Tony: keyword briefs, SEO/GEO drafts, quality audit, preflight, and source-ready delivery. Tony does not claim live production publication.
Factory: creates the daily 12-post SEO/GEO package, phase receipts, manifests, repairs, and local publish artifacts.
Peter: ClawLite publish apply, build QA, optional live deploy, live QA, canonical safety, and closeout.
Patrol: post-publish and daily visibility checks for SERP/indexing, llms.txt, GA4, canonical/cannibalization, locale paths, and rescue signals.

Required Phase Matrix

Every generated article should have receipts for:

Research: keyword-research, competitor-analysis, serp-analysis, content-gap-analysis
Build: seo-content-writer, geo-content-optimizer, meta-tags-optimizer, schema-markup-generator
Optimize: on-page-seo-auditor, technical-seo-checker, internal-linking-optimizer, content-refresher
Monitor: rank-tracker, backlink-analyzer, performance-reporter, alert-manager
Cross-cutting: content-quality-auditor, domain-authority-auditor, entity-optimizer, memory-management

Monitor receipts may be NEEDS_DATA until live URLs, rank provider data, analytics, and backlink data exist.

Receipt Map

Core daily receipts:

mission-control/data/runner/hunter-tony-seo-geo-YYYY-MM-DD.json
mission-control/data/runner/hunter-daily-topic-delivery-YYYY-MM-DD.json
mission-control/data/runner/clawlite-publish-apply-build-YYYY-MM-DD.json
mission-control/data/content-delivery/hunter-daily-topic-radar-raw-YYYY-MM-DD.json
mission-control/data/content-delivery/hunter-daily-topics-YYYY-MM-DD.json
mission-control/data/content-delivery/tony-keyword-brief-YYYY-MM-DD.json
mission-control/data/content-delivery/tony-content-quality-audit-YYYY-MM-DD.json
mission-control/data/content-delivery/tony-blog-preflight-YYYY-MM-DD.json
mission-control/data/content-delivery/tony-blog-source-publish-YYYY-MM-DD.json
mission-control/data/peter/peter-blog-closeout-verify-YYYY-MM-DD.json
mission-control/data/peter/peter-live-publish-qa-YYYY-MM-DD.json
mission-control/data/seo-patrol/gr-seo-patrol-YYYY-MM-DD.json
mission-control/data/seo-patrol/gr-seo-patrol-YYYY-MM-DD.md

Daily package outputs:

delivery/seo-geo/YYYY-MM-DD/manifest.md
delivery/seo-geo/YYYY-MM-DD/manifest.json
delivery/seo-geo/YYYY-MM-DD/blogs/
delivery/seo-geo/YYYY-MM-DD/metadata/
delivery/seo-geo/YYYY-MM-DD/audits/
delivery/seo-geo/YYYY-MM-DD/skills/\x3Cphase>/\x3Cskill>/
delivery/seo-geo/YYYY-MM-DD/repairs/
delivery/seo-geo/YYYY-MM-DD/publish/clawlite-content/
synthadoc/wiki/*hunter-daily-topic*

Operating Loop

Identify target date. Default to today only when Ray did not provide a date.
Read the runner receipt first: mission-control/data/runner/hunter-tony-seo-geo-YYYY-MM-DD.json.
If the runner receipt is missing and Ray asked to run delivery, run the full daily command.
Inspect Hunter raw radar and durable topic delivery before judging topic selection.
Inspect delivery/seo-geo/YYYY-MM-DD/manifest.md and manifest.json before judging article readiness.
Check repairSummary, skillPipeline, publishQueue, deploy, patrol, and Peter closeout.
If any article has humanReviewRequired > 0, inspect repairs/ and failed skill receipts before editing.
If Peter/local build passed but live deploy was skipped, say staged/build passed, not published.
If live deploy ran, verify deploy evidence and peter-live-publish-qa.
Treat overall workflow as not fully green until patrol blockers are cleared.

Status Meanings

READY: publish-queue eligible.
BLOCKED: at least one blocking gate failed.
PASS: a required step completed successfully.
FAIL: a blocking step failed and needs repair or human review.
NEEDS_DATA: non-blocking monitor step is waiting for live URL or connector data.
PASS_WITH_WARNINGS: delivery may continue, but connector data or non-blocking checks are missing.
STAGED_LOCAL: source files or local repo changes were prepared; production publication is not proven.
SKIPPED: a step intentionally did not run and must include a reason.

Gates

An article can enter publish queue only when:

all blocking Research, Build, Optimize, and Cross-cutting receipts are PASS
draft audit status is PASS
audit score is at least 90
repair status is NOT_NEEDED or REPAIRED
no blocking skill receipt has FAIL

If a blocking gate fails:

Mark the article BLOCKED.
Map audit failures to failed skills.
Write delivery/seo-geo/YYYY-MM-DD/repairs/\x3Carticle>.json.
Repair only the failed surface: structure, word count, keyword placement, internal links, meta, or schema.
Rerun failed skill and downstream receipts.
Queue only if final audit reaches PASS and score >= 90.
If repair is disabled or still fails, mark failed blocking receipts FAIL with humanReviewRequired: true.

Common Diagnostics

"The blog did not publish"

Check in order:

mission-control/data/runner/hunter-tony-seo-geo-YYYY-MM-DD.json
delivery/seo-geo/YYYY-MM-DD/manifest.md
mission-control/data/runner/clawlite-publish-apply-build-YYYY-MM-DD.json
mission-control/data/peter/peter-blog-closeout-verify-YYYY-MM-DD.json
mission-control/data/peter/peter-live-publish-qa-YYYY-MM-DD.json
Live URL and sitemap only after Peter live QA exists.

Report the exact stage: Hunter, Tony, Factory, Peter local apply/build, live deploy, live QA, or patrol.

"Did you use the SEO/GEO skill?"

Check delivery/seo-geo/YYYY-MM-DD/skills/ and manifest.json. The expected skill family is seo-geo-claude-skills, with article-level receipts for the phase matrix. Do not answer from memory.

"Daily workflow is stuck"

Check:

OpenClaw logs: /Users/m1/.openclaw/logs/*.log
LaunchAgents: launchd/ai.openclaw.daily-seo-geo-blog-factory.plist
runner receipt freshness under mission-control/data/runner/
active OpenClaw sessions if the task was triggered through Feishu/Discord/Telegram

If a chat-triggered task timed out, use the receipts to determine whether work completed after the chat reply failed.

Patrol blockers

Known examples:

GA4_TAG_MISSING: Peter/build/live site issue. Verify built layout and live HTML.
DATAFORSEO_B64_MISSING: connector/env issue. Configure env/Keychain before rank/index checks can be complete.
locale /en/, /ja/, /ko/ 404 or redirect warnings: usually non-blocking unless Ray wants locale pages live.
no top-100 keyword hits: monitor finding, not a publish blocker by itself.

Production alias safety

If clawlite.ai shows old Mission Control or /mission-control unexpectedly returns 200, suspect the legacy Vercel project clawlite reclaimed the alias. Restore to the verified safe ClawLite main deployment before claiming live success.

Patrol and Rescue Routing

new top-30 long-tail opportunity -> Hunter
title/body/internal-link decay -> Tony
canonical, live URL, build, deploy, alias, sitemap, or GA4 issue -> Peter
missing connector or API credential -> environment/config maintenance

Every daily run should produce a gr-seo-patrol receipt. Patrol checks include SERP rank diff when DATAFORSEO_B64 exists, Google index count when supported, llms.txt, GA4, canonical/cannibalization, rescue signals, rank-tracking slots, and redirect-aware locale checks.

Related Skills and References

ray-tony-seo-geo: use for single-article Tony keyword/content/audit/source-publish work.
gr-seo-patrol: use for focused patrol/rescue work.
/Users/m1/seo-geo-claude-skills: upstream SEO/GEO phase skill library.
references/clawlite-seo-geo-workflow.md: workflow reference notes.

安全使用建议

Install this only for the intended ClawLite/OpenClaw environment. Before running it, verify the local scripts it references, keep provider credentials scoped, and require explicit approval for any command that writes to the main site or deploys to production.

功能分析

Type: OpenClaw Skill Name: seo-geo-workflow-for-clawlite Version: 0.4.3 The skill bundle defines a complex SEO/GEO workflow that requires the AI agent to execute local Node.js scripts and Vercel deployment commands within a specific user directory (/Users/m1/). While these capabilities are clearly aligned with the stated goal of automating content production and monitoring, the reliance on shell execution and broad file system access represents a high-risk capability according to the analysis criteria. No evidence of malicious intent, data exfiltration, or obfuscation was found in SKILL.md or the associated files, but the high-privilege nature of the workflow commands warrants a suspicious classification.

能力评估

ℹ Purpose & Capability

The stated purpose—Hunter/Tony/Peter SEO-GEO content production, QA, and patrol—matches the commands and receipt workflow described. The capability is high-impact because it includes content production, publish readiness, and guarded live deployment paths.

ℹ Instruction Scope

The instructions include meaningful guardrails: receipt-first verification, no false production claims, no main-site writes without an explicit override, and live deploy only with deploy evidence. Users should still approve high-impact commands before execution.

ℹ Install Mechanism

There is no install spec and no bundled code; the skill depends on local scripts under /Users/m1/.openclaw/workspace. That is not suspicious by itself, but those scripts are outside this review and should be verified separately.

ℹ Credentials

The absolute local paths and optional DataForSEO, GA4, and deployment-related environment values are proportionate for Ray's ClawLite/OpenClaw workflow, but they are user-specific and not portable.

ℹ Persistence & Privilege

The skill relies on durable receipts and daily package outputs, which fit the workflow. It does not show hidden background persistence, but explicit main-site writes or live deploys could affect a public site.

版本历史

v0.4.3

Document protected main-site write guard and explicit override requirement

v0.4.2

Add SEO README and keyword-focused usage overview

v0.4.1

Rename and publish controlling runbook for OpenClaw daily SEO-GEO Hunter/Tony/Peter workflow

元数据

Slug seo-geo-workflow-for-clawlite

版本 0.4.3

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 3

常见问题