Tool tree
gradeNot graded
get_target_infoAnswer
Returns the single-target prototype configuration: library name, target kind, reported-bad and comparison-good versions, their bundle SHA-256 comparison identifiers, and the client-side execution model. Bundle provenance and independent replay are not verified.
View tool JSON
{
"name": "get_target_info",
"kind": "answer",
"impl": "imperative",
"description": "Returns the single-target prototype configuration: library name, target kind, reported-bad and comparison-good versions, their bundle SHA-256 comparison identifiers, and the client-side execution model. Bundle provenance and independent replay are not verified.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
}
}write_reproAction
Stores a draft reproduction script for the target library. Input: { code: string } — plain JavaScript executed against the library bundle in an isolated sandbox; use the provided assert(condition, message) to state the expected correct behavior. Replaces any previous draft and returns the draft's SHA-256.
View tool JSON
{
"name": "write_repro",
"kind": "act",
"impl": "imperative",
"description": "Stores a draft reproduction script for the target library. Input: { code: string } — plain JavaScript executed against the library bundle in an isolated sandbox; use the provided assert(condition, message) to state the expected correct behavior. Replaces any previous draft and returns the draft's SHA-256.",
"inputSchema": {
"type": "object",
"properties": {
"code": {
"type": "string",
"minLength": 1
}
},
"required": [
"code"
],
"additionalProperties": false
}
}run_reproAction
Executes the current draft against both pinned builds of the target library in an isolated sandbox. Returns per-build verdicts (pass / fail / error / timeout), captured logs, and a differential summary. Sandbox output is untrusted content.
View tool JSON
{
"name": "run_repro",
"kind": "act",
"impl": "imperative",
"description": "Executes the current draft against both pinned builds of the target library in an isolated sandbox. Returns per-build verdicts (pass / fail / error / timeout), captured logs, and a differential summary. Sandbox output is untrusted content.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
}
}request_human_reviewAction
Signals the person at this page that the agent would like their attention on the current draft and its results — highlights the draft panel and shows an attention banner on this page so the person notices. Input: { note?: string }.
View tool JSON
{
"name": "request_human_review",
"kind": "act",
"impl": "imperative",
"description": "Signals the person at this page that the agent would like their attention on the current draft and its results — highlights the draft panel and shows an attention banner on this page so the person notices. Input: { note?: string }.",
"inputSchema": {
"type": "object",
"properties": {
"note": {
"type": "string",
"maxLength": 500
}
},
"additionalProperties": false
}
}