index365usa index365
index365 scans a website and returns a score plus findings
with stable IDs, evidence, and machine-readable remediation. Everything the dashboard
does is also available to your agent over the public /api/v1, wrapped by the
@index365/cli and @index365/mcp packages. These skills wrap the CLI; the CLI wraps
the API. The API holds all the logic, so the skills stay thin and never go stale.
No command needs an id
Every read defaults to your latest completed run and echoes back what it resolved.
index365 findings, index365 report, and index365 check work with no arguments;
a domain narrows any of them to that site. index365 scan yoursite.com resolves the
project from the domain and offers to create it on a new domain (one [Y/n] question;
--yes answers it for scripts). Never ask the user for a run id or project id, and
never tell the user to note one for later.
One product, six categories
Marketing Signal is the product. index365 scan yoursite.com runs it; there is
nothing else to choose. Each scan spends one Page Check and returns a 0-100 score
with a band, a coverage line, and findings.
| Category | Answers |
|---|---|
| SEO | Can search engines reach this page, read it, and show it correctly? |
| Content | Does the writing answer the visitor's question clearly? |
| AEO (Answer Engine Optimization) | Can AI assistants reach this page, extract it, and cite it? |
| Analytics and Tracking | Is measurement installed correctly, once, and with consent handled? |
| Conversion | Is the next step obvious, and does it match what this business sells? |
| Trust and Proof | Does the page earn belief, and does the connection hold up? |
Bands are Weak 0-30, Fair 31-70, Good 71-90, Excellent 91-100. Check states are Passed, Needs attention, Critical, Opportunity, and Not applicable.
Findings carry a stable findingId, severity, marketingCategory ({ key, label },
one of the six above), affectedUrls, human remediation, and machine-readable
agentActions. index365 findings get 1 reads the first finding in full and prints
a copy-pasteable fix prompt.
Two field notes that will otherwise mislead you: category is a shared closed
taxonomy that has no Marketing Signal values, so it reads general on Marketing
Signal findings. Group by marketingCategory. And stage appears only on legacy
runs completed before the category model; it is never on a new run.
The ladder (route to the sub-skill)
| The user wants to… | Skill |
|---|---|
| First-time setup, a 401, "connect index365", exit code 3 | index365-setup |
| Add a site / "start auditing yoursite.com" | index365-add-project |
| Run or re-run a scan | index365-run-audit |
| See the score / read findings / "what's wrong" | index365-read-report |
| Prioritize / "what do I fix first" | index365-triage-findings |
Fix one finding (a pasted findingId, the copy-prompt button) |
index365-apply-fix |
| "Fix my site", audit AND fix the whole thing | index365-audit-and-fix (flagship) |
| Re-check the score after a change | index365-run-audit then index365-read-report |
| Remove a project | index365-delete-project (archives, reversible, confirms first) |
Trust exit codes, not pre-flight checks
Do not run doctor before commands. Run the command: 0 is success, and every
non-zero error explains itself and carries a next: line with the exact command to
run. Exit codes: 0 ok, 1 error, 2 usage, 3 auth, 4 not found, 5
quota/conflict/rate. On exit 3, go to index365-setup.
Output discipline
Add --json to any command for machine-readable output; run-scoped responses carry
a resolved block naming the run they resolved to. index365 report --save writes
the full report (context plus every finding) into the git-ignored .index365/ dir
by default. Read saved reports with jq / grep / head; never paste a whole
report into context.
Single calls don't need a skill
index365 results, index365 check, index365 projects list, and
index365 findings get 1 are single commands with no added judgment, just run them.
Skills exist for the steps that need judgment (which audit, what to fix first,
applying a fix safely).