# Refusal Desk — pre-execution task-brief tribunal (P58) — TextForAgents LLC
Fleets burn budget executing briefs that were doomed at intake. The Refusal Desk screens each incoming brief before execution — underspecification, internal contradiction, standing-instruction conflict, semantic duplication of completed fleet work — and returns a priced verdict: ACCEPT (defect classes cleared, with a failure forecast), REPAIR (a structured brief diff, never a silent rewrite), REFUSE (cause code priced against the execution prevented), or DUPLICATE-OF with the prior result attached. Forecasts are settled against downstream attributed outcomes; calibration is published per defect class and a forecast miss is priced at $0.05, capped at $5.00 per principal per period.
**Per-check $0.01** (x402) · **Fleet $19/mo** (up to 5,000 checks/mo) · **Duplicate lookup +$0.02/lookup** · First 100 checks free, one time per operator.
Docs/packet summary: refusal-desk-task-brief-tribunal — see the product page for the full description, limits, and support (support@textforagents.com).
Product: https://textforagents.com/products/refusal-desk-task-brief-tribunal/
API: POST /register (unauthenticated) -> account_id + claim token; POST /trial (claim token) -> API token + 100 free checks; POST /check_brief {brief} ($0.01 / trial / Fleet quota); POST /duplicate_lookup {brief} (+$0.02); POST /settle_outcome {verdict_id, outcome, failure_class?, result?, cost_usd?}; GET /calibration_report; POST /standing-rules (P45 compiled-entry shape); POST /p45-link {ledger_id, p45_token} to bind the account to a deployed P45 principal-drift ledger, consulted live on every check (GET /p45-link for link + last-consult status; POST {unlink: true} to unbind); POST /subscribe (Fleet $19/mo via x402); POST /mcp (JSON-RPC: check_brief, settle_outcome, duplicate_lookup, calibration_report).
Limits: Documented limits: the Desk is deterministic and LLM-free — it screens structure, contradictions, standing-instruction conflicts, and semantic duplication against the account's own fleet history; it does not judge whether a well-formed brief is wise. Duplicate detection is semantic-similarity based over the account's own completed-work history for the same principal (work settled as successful through this Desk, with or without a result text), is period/number sensitive (Q3 work is not Q4 work), and can miss heavily rewritten duplicates. Standing-instruction conflicts are checked against the account's stored standing rules in the P45 compiled-entry shape and — for accounts that link a P45 ledger — against that ledger consulted live on every check (its active compiled prohibitions; other polarities are counted on the verdict, not enforced here); natural-language rules beyond the compiled pattern shape are not interpreted. The share of fleet spend lost to defective briefs is UNMEASURED — the lab census found the Lab's own internal briefs only 3.5% intake-defective (brief-first discipline), which cannot validate external demand; no defect-rate or savings figure for real fleets is claimed.
Support: support@textforagents.com