Repository navigation
BeyondSEO 2.11.0 makes one request do the whole job. Ask for an audit and your assistant now has clear steps to follow, a way to check nothing was skipped, and a report you can actually read.
What's new
- Complete audits. The skill now gives your assistant exact steps: review the business, compare real competitors, check backlinks and write a 30/60/90-day plan.
audit-check. This new command compares the audit folder and report with the required stages. It flags missing evidence, names and counts that do not match, and claims that are not attributed to a source.- Reports people can read. Repeated issues become one card, reports open with your top actions, and evidence appears in plain words. In one test, the same saved audit went from 142 pages to 22.
audit.jsonstill keeps every observation. - Clear warnings. A website-only export now says it is not a complete audit.
- A new README and explainer. See what BeyondSEO does in about 20 seconds, then install from one row of buttons.
- Smaller fixes. Curly quotes now match in evidence, quote and manifest errors explain what to do, and the guide covers Codex's sandbox limit on Chromium.
Install or update
Download beyondseo-skill.zip below for apps that accept a skill upload. beyondseo-2.11.0-skill.zip is the same file under a versioned name, and each has a .sha256 checksum. Use the host-specific installation guide. GitHub's automatic source ZIP is not a direct skill upload.
An older copy can take priority over a newer one, so ask your assistant which version and path it loaded after updating.
Validation and limits
- Local regression suite on macOS with Python 3.10, 3.12 and 3.13: 357 tests passed, one Windows-only test skipped.
- All GitHub CI jobs passed across Linux, macOS and Windows with Python 3.10 and 3.12 on this release commit.
- Ruff, formatting, documentation links, package format and release hygiene checks passed. The upload has 200 files including its receipt, within the observed hosted upload limit.
- Live full audits of one real website, run just before the last small changes in this release: Codex reviewed 15 candidate competitors, selected 5 direct competitors, checked 30 sources and compared 6 sites;
audit-checkrated it complete with limits because rendering was blocked in its sandbox and rankings were withheld. Claude Code reviewed 7 competitor sites and 7 sources, also complete with limits. - Live checks of this exact release: Claude Code and Codex loaded 2.11.0, crawled a page and ran
audit-check. Hermes loaded 2.11.0 (read-only check). - Not run end to end with this release: ChatGPT Work, Claude Cowork, the Claude website, Cursor, Kimi Code, the Kimi website and OpenClaw. The installer was tested for Cursor, Kimi Code and OpenClaw in clean test folders. Tool permissions, authentication and network access depend on the host.
audit-checkconfirms that evidence exists and that key names and counts reach the report. It does not judge whether the analysis is correct.- Readiness is not a guarantee of indexing, ranking or AI citations.
See the changelog.