Site fixes
site_fixesAudit the live site the way AI crawlers and Google read it (robots rules for AI crawlers, firewall, rendering, indexing, canonicals, page titles and descriptions, structured data, sitemap, llms.txt, internal links to key pages, share images, answer pages) and return fix specs { id, theme, page, issue, evidence, fix } grouped by theme, one PR per theme, minus fixes already shipped. Costs $0.02 from the balance, charged only when it returns. Nothing is charged when it finds nothing to fix. When you record the PR, put the fix ids in meta.fixIds so they are not suggested again. The robots fix is the whole corrected robots.txt, with a note on each crawler it changes for the PR. Training crawlers (GPTBot, ClaudeBot, CCBot...) are the founder's call: while they have not answered, that fix carries fix.choices, the whole file for allow, block and keep; ask, use the one they pick, and put their answer in the PR's record as meta.training so later checks follow it.
Tool call template
This tool declares no required parameters. Add optional parameters from the input schema as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {},
"name": "site_fixes"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| url | No | A page on this workspace's site. Default: the homepage | |
| training | No | The founder's answer on training crawlers: allow (let them in), block (keep them out), keep (leave robots.txt as it is for them). Kept for later checks; left out, the last answer stands |