{"name": "In-situ prompt-injection detector verdict", "product": "P167", "version": "1.0.0", "provider": "TextForAgents LLC", "one_liner": "Your detector's public benchmark score does not predict how it behaves inside YOUR agent - grade it on YOUR tool outputs: catch at 1% FPR, per surface, with a training-overlap audit, in one signed receipt.", "what_it_does": "Grades a prompt-injection detector (or a shortlist) on the buyer's own tool-output distribution: differential replay labels injected outputs as twins of benign outputs with the payload buried at a recorded position across the buyer's tool surfaces; the detector is calibrated to a fixed 1% benign-FPR operating point; the signed verdict reports catch@1%FPR, benign FPR, and tasks-stopped rate per surface, plus a training-data overlap audit, pinned to detector id x version x distribution hash, so a version bump or a new distribution reads STALE, never silently current.", "limits": "Lab figures (BIPIA leader 0.1933 in-situ vs 0.97 claimed; tau -0.3333) were measured on the lab's synthetic distribution against detector test doubles replicating documented failure shapes - they demonstrate the mechanism, not a rating of any named production detector. Whether a synthetic-but-buyer-shaped distribution grades close enough to a real recorded one is unmeasured in the wild. A detector with undisclosed training data grades UNDISCLOSED. Willingness to pay is unmeasured. A grade is not a live-attack guarantee.", "prices": {"grade_lookup_per_query_usd": 0.01, "grade_classifier_usd": 4.99, "grade_judge_usd": 9.99, "shortlist_comparison_usd": 19.99, "regrade_pass_per_month_usd": 9.99, "trial": "one free in-situ grade (classifier detector) per operator, ever - bound to the account and the payer wallet declared at registration"}, "fences": "Shortlist $19.99 is the comparison price for a shortlist of up to 5 detectors and covers the classifier detectors in it; every judge detector inside a shortlist meters separately at $9.99 in addition. The re-grade pass covers one named classifier detector's re-grades plus version-bump alerts for the operator's other named detectors; additional re-grades meter per grade. A grade is the standard sample block (up to 1350 outputs); larger samples meter per block. A buyer-named detector endpoint must be https on a public, routable host; redirects are not followed and responses over 256 KB are refused.", "how_to_start": "POST /register {name, payer_wallet} -> claim_token; POST /claim {claim_token} -> your api token (shown once). Send it as Authorization: Bearer <token>. Your first classifier grade is free (trial). Paid calls answer 402 with an x402 challenge; pay it and resend with the X-PAYMENT header - payment verifies through the facilitator before work and settles after.", "endpoints": {"POST /grade": "grade one detector (built-in detector_id or buyer detector_spec with its own https endpoint)", "POST /compare": "rank a shortlist of up to 5 detectors on one distribution", "POST /verify": "verify a signed receipt (free)", "GET /lookup": "latest signed grade pin check for one detector ($0.01; version-pin STALE answers are free)", "POST /subscribe": "start the $9.99/mo re-grade pass for one named classifier detector", "GET /envelope": "your account: trial, pass, receipts", "POST /samples/delete": "tombstone your grade records for a distribution hash", "GET /detectors": "built-in detectors and their claimed public scores", "GET /health": "service health"}, "support": "support@textforagents.com"}