seo-engine: day 11 PM and day 12 AM, Agent SDK, agent teams, ZCode, the 9 Oct brief

[chaitanyagiri]

What changed

Three new pages and one fix, on one branch so there is one merge.

• /blog/what-is-claude-agent-sdk/: what the SDK is, languages, billing and limits. "claude agent sdk" has about 12,300 searches a month at difficulty 4 (OpenSEO planning sprint). 1,154 words, four scenes (musicbox, penknife, points, eggbox). One link from local-first-vs-cloud-agent-sdks.
• /blog/claude-code-agent-teams/: the experimental agent teams feature, from Anthropic's docs. "claude code agent teams" has about 2,900 searches a month at difficulty 4. 1,147 words, four scenes (oars, switchboard, tickets, meters). One link from manage-multiple-claude-code-sessions.
• /blog/what-is-zcode/: Z.ai's coding agent harness. "zcode" has about 9,900 searches a month at difficulty 1. 1,107 words, four scenes (fan, dial, abacus, dominoes). One link from best-ai-coding-agents.
• Fix: does-the-june-2026-agent-sdk-change-affect-munder-difflin said the June 15 Agent SDK billing change took effect. Anthropic's support article says it was paused on June 15, 2026. Added a dated update box at the top, and corrected the description and the three FAQ answers. The body is left as written and marked as the change as announced.

Checks

• Gate: pass on the three new pages with media and motion checks on. Build and node tools/check-release-links.cjs: pass.
• Independent fact review, each reviewer opening every source on 8 Oct 2026:
  • what-is-zcode: pass on round 1; five wording advisories applied.
  • claude-code-agent-teams: round 1 failed on one dropped qualifier in a FAQ answer ("but practical constraints apply"); fixed; round 2 by a different reviewer passed.
  • what-is-claude-agent-sdk: round 1 failed on one caption that dropped "on Max and Team plans"; fixed; round 2 by a different reviewer passed.
  • The correction box, description and FAQ sentences on the June post: checked against Anthropic's support article by the round 2 reviewer, pass. I also read the article myself.
• None of the pages claims we tested the tool. The ZCode page says so and marks text taken from the repo's Chinese NOTICE.md as our translation.
• Every number and label shown in the scenes was checked against the post body by me; I changed two labels that were not in the body and reshot those stills. Scenes swap in on the built pages in headless Chrome (4 of 4 on each).
• Not tested: tap and keyboard actions in the scenes; Safari, Firefox, a real phone.
• Known and left as is: the three older posts that received a link fail the gate on length, dashes or headings before this change too. Only one sentence was added to each. Two title cards in this pull request both use kites.
• The June post is a founder authored post about the product. I did not check its claims about the app (the enrich assistant and claude -p); I only corrected the Anthropic billing facts.

Before

The three new URLs returned 404 on 8 Oct 2026, and the June post had no correction:

After

Added 9 Oct, morning run (commit 4d20134)

• New page /blog/agent-tools-today-2026-10-09/: the daily brief, 17 items and 4 talked about, one scene. Independent fact review passed (78 claims opened against their sources); four small wording fixes applied after it.
• is-ollama-good-for-coding: clearer title and description only.
• Not in this PR: what-is-gemini-agent. It failed fact review three times and is held in the backlog.
• Not checked: the scene's tap interaction in a browser (still frames only).

Before

After

🤖 Generated with Claude Code