gradeNot graded
feedDocumentAction
Start here. Pass the article the user gave you: url, plus requested_terms if they named any words to look up (omit it if they did not — then judge for yourself what is worth reporting). Returns the extraction rules, the finding schema and a thin lexicon index. Read-only; the page never fetches the article, you do. THEN DO THIS, IN ORDER: (1) open the article; (2) call reportDocument once with its title, byline, date, rough length and a one-sentence gist — the person is staring at a blank screen while you read, and this is the only thing that tells them you got there; (3) extract the terms and send them with submitFindings.
View tool JSON
{
"name": "feedDocument",
"kind": "act",
"impl": "imperative",
"description": "Start here. Pass the article the user gave you: url, plus requested_terms if they named any words to look up (omit it if they did not — then judge for yourself what is worth reporting). Returns the extraction rules, the finding schema and a thin lexicon index. Read-only; the page never fetches the article, you do. THEN DO THIS, IN ORDER: (1) open the article; (2) call reportDocument once with its title, byline, date, rough length and a one-sentence gist — the person is staring at a blank screen while you read, and this is the only thing that tells them you got there; (3) extract the terms and send them with submitFindings.",
"inputSchema": {
"type": "object",
"properties": {
"url": {
"type": "string"
},
"requested_terms": {
"type": "array",
"items": {
"type": "string"
}
}
},
"required": [
"url"
],
"additionalProperties": false
}
}reportDocumentAction
Call this once, right after you open the article and before you start extracting terms. Report what you can see of the document: title (verbatim, required), byline, published (YYYY-MM-DD, only if the page states one), words (rough length), gist (one sentence, max 100 chars, on what the article is about). The person is watching a blank screen while you read; this is what tells them you reached the page. Nothing here is stored.
View tool JSON
{
"name": "reportDocument",
"kind": "act",
"impl": "imperative",
"description": "Call this once, right after you open the article and before you start extracting terms. Report what you can see of the document: title (verbatim, required), byline, published (YYYY-MM-DD, only if the page states one), words (rough length), gist (one sentence, max 100 chars, on what the article is about). The person is watching a blank screen while you read; this is what tells them you reached the page. Nothing here is stored.",
"inputSchema": {
"type": "object",
"properties": {
"title": {
"type": "string"
},
"byline": {
"type": "string"
},
"published": {
"type": "string"
},
"words": {
"type": "number"
},
"gist": {
"type": "string"
}
},
"required": [
"title"
],
"additionalProperties": false
}
}submitFindingsSensitive Action
Send the AI-term sightings found in the article back to the dictionary. This does NOT write to the dictionary: it returns status pending_review and the findings wait on the page until the person confirms them. That is the final answer to this call — do not resubmit or poll.
View tool JSON
{
"name": "submitFindings",
"kind": "transact",
"impl": "imperative",
"description": "Send the AI-term sightings found in the article back to the dictionary. This does NOT write to the dictionary: it returns status pending_review and the findings wait on the page until the person confirms them. That is the final answer to this call — do not resubmit or poll.",
"inputSchema": {
"type": "object",
"properties": {
"findings": {
"type": "array",
"items": {
"type": "object",
"additionalProperties": false,
"required": [
"term_raw",
"term_normalized",
"sentence",
"context",
"explained",
"intent",
"domain",
"definition_quote",
"requested"
],
"properties": {
"term_raw": {
"type": "string",
"description": "exactly as written in the document"
},
"term_normalized": {
"type": "string",
"description": "the lexicon term this maps to; \"\" if not a known term"
},
"sentence": {
"type": "string",
"maxLength": 120,
"description": "the one sentence it appears in, de-identified"
},
"context": {
"type": "string",
"description": "sentence plus one sentence before and after, de-identified"
},
"explained": {
"type": "string",
"enum": [
"has_definition",
"mentioned",
"assumed"
]
},
"intent": {
"type": "string",
"enum": [
"selling_point",
"technical",
"risk_or_limit"
]
},
"domain": {
"type": "string",
"enum": [
"core",
"edge",
"not"
]
},
"definition_quote": {
"type": "string",
"description": "verbatim substring of context; \"\" if the document does not define the term"
},
"requested": {
"type": "boolean",
"description": "true if the user listed this term in requested_terms; false if you added it"
},
"source": {
"type": "object",
"additionalProperties": false,
"description": "where the article came from; url from feedDocument",
"properties": {
"url": {
"type": "string"
},
"title": {
"type": "string"
},
"published": {
"type": "string",
"description": "YYYY-MM-DD or \"\""
}
},
"required": [
"url"
]
}
}
}
},
"not_found": {
"type": "array",
"items": {
"type": "string"
}
}
},
"required": [
"findings",
"not_found"
],
"additionalProperties": false
}
}lookupTermAnswer
Ask the dictionary what a term currently looks like on the web: every definition quote with its source. Evidence only; the dictionary does not rule which definition is right. Read-only.
View tool JSON
{
"name": "lookupTerm",
"kind": "answer",
"impl": "imperative",
"description": "Ask the dictionary what a term currently looks like on the web: every definition quote with its source. Evidence only; the dictionary does not rule which definition is right. Read-only.",
"inputSchema": {
"type": "object",
"properties": {
"term": {
"type": "string"
}
},
"required": [
"term"
],
"additionalProperties": false
}
}trendingAnswer
Top terms by sightings in the last 30 days. Read-only.
View tool JSON
{
"name": "trending",
"kind": "answer",
"impl": "imperative",
"description": "Top terms by sightings in the last 30 days. Read-only.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
}
}