seo-run
Skill1 starsChanged 4 months agoMITView on GitHub
Use as the single entry point of the SEO.md system, the command to invoke on a schedule (cron, GitHub Action, scheduled agent). Decides and executes exactly one content action per run.
In this repository5 other files it gives its agents.
- seo-initSkill
- seo-planSkill
- seo-programmaticSkill
- seo-toolSkill
- seo-writeSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
Read SEO.md at the project root, in full.
- Distinctive: line 13
Never invent product or audience facts that are not in SEO.md.
- Very specific: line 14
Default paths (overridable in SEO.md § State): backlog .seo/backlog.json, decision log .seo/log.md.
- Distinctive: line 44
If the invoked skill's quality gate fails, report the failure and STOP.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
seo-run: decide and run the next best SEO action
Before anything
- Read
SEO.mdat the project root, in full. It is the only source of truth for product, audience, voice, rules, and paths. If it does not exist: stop and tell the user to run the seo-init skill (or fillSEO.template.mdby hand). - Never invent product or audience facts that are not in SEO.md.
- Default paths (overridable in SEO.md § State): backlog
.seo/backlog.json, decision log.seo/log.md.
Decision procedure (deterministic, in order)
- Read state: SEO.md, the backlog, and the decision log.
- Cadence guard: find the most recent
publishedAtacross backlogdoneentries and published content. If it is today, STOP: "already published today". Never double-publish a day. Also respect SEO.md § Content Mix & Cadence (e.g. at "3 per week", if 3 pieces shipped in the last 7 days, stop and say when the next slot opens). - Backlog health: count
todoentries. Under 20: invoke the seo-plan skill, then STOP for this run (refreshing the backlog IS the action; one action per run). - Inventory the published mix: count published pieces by
type(backlogdoneentries plus anything obvious in the content locations).glossarycounts in the programmatic bucket: the mix has three buckets (articles / tools / programmatic). Compute each bucket's actual share. - Pick the type: the type furthest BELOW its target share in SEO.md § Content Mix
& Cadence that has at least one
todotopic available. Tie-break: lowerprioritynumber first, then higher opportunity (strong intent, high volume, low difficulty). Worked example: targets 70/15/15, published 12 articles + 1 tool + 0 programmatic (shares 92%/8%/0%): programmatic is furthest below target; if no programmatic topic istodo, fall back to tool; if none either, article. - Invoke the matching skill via the Skill tool (or your agent's skill
mechanism):
seo-write(article),seo-tool(tool),seo-programmatic(programmatic or glossary). Exactly one. - Log the decision: append one line to
.seo/log.md:2026-07-07 | article | reduce-return-rate-shopify | articles 62% vs 70% target, top priority todo
Failure handling
If the invoked skill's quality gate fails, report the failure and STOP. Do not retry with a different topic in the same run: a failed gate is signal about the backlog or the rules, and burning through topics hides it.
Never do
- Two content actions in one run (a seo-plan refresh counts as the run's action).
- Bypassing a worker skill or its gate "because the decision is obvious".
- Inventing a topic when the backlog has none of the needed type: that is seo-plan's job.
- Editing SEO.md: the contract changes only via seo-init or the user's hand.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

