Check a text
check_textDecide whether one text should be allowed: profanity, slurs, insults, threats and sexual language, disguised spellings included (sh1t, f.u.c.k, leetspeak, homoglyphs). Returns the action (allow, review, block), the matches with their categories and the tricks used, a one-sentence reason and a censored version. Read-only; counts one check against the plan. If the action is wrong, report it with report_result and the result's id.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"text": "<text>"
},
"name": "check_text"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| mask | No | How "censored" hides a matched word: asterisk (****), first_letter (f***), grawlix (#$@!) or remove. Omit for the project's default, asterisk unless the project set one. | |
| mode | No | How much to ask the model: fast never asks the model; balanced, the default, asks it about every text but a word or two the rules block outright; thorough asks it about every text. Omit for the project's default. | |
| text | Yes | The text to check, exactly as the person typed it: up to 10,000 characters, 64 for a username or display name. | |
| policy | No | How strict to be. A preset: kids (under 13, blocks everything), teen, standard (adults; the default), strict (brand-safe surfaces), casual (swearing welcome, swearing at someone still blocked). Or a saved policy id (pol_...), or an object giving an action per category. Omit for the project's default. | |
| context | No | Where the text is shown, which changes how it is read: username (a public handle; reserved names match), display_name (a name beside content), comment, chat, gaming (taunts about play are not abuse), review, educational (quoted words are discussed, not used), creative (dialogue in fiction), custom_text (engraving or printing), general (anything else). Omit for the project's default, general unless the project set one. | |
| country | No | Where the readers are, as a two-letter country code (MX, ES, BR): a word rude in some countries and innocent in others is read as theirs reads it. Omit to count a word rude anywhere. | |
| language | No | The language the text is in, when you know it: a short text that shows no other language is read in it. Defaults to auto. |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | ||
| score | Yes | ||
| action | Yes | ||
| engine | Yes | ||
| reason | Yes | ||
| allowed | Yes | ||
| evasion | Yes | ||
| matches | Yes | ||
| censored | Yes | ||
| language | Yes | ||
| severity | Yes | ||
| timing_ms | Yes | ||
| categories | Yes | ||
| decided_by | Yes | ||
| obfuscated | Yes |