01 Prohibited categories
No skills that extract, capture, exfiltrate, or misuse passwords, API keys, session cookies, tokens, cards, private keys, or other secrets.
No skills that secretly copy, package, upload, or export user files, browser data, source code, private documents, screenshots, messages, or other local data to external endpoints.
No keyloggers, reverse shells, backdoors, persistence, evasion, unauthorized implants, or instructions to bypass security tools.
No skills that wipe disks; delete or corrupt credential stores, browser profiles, source trees, workspaces, backups, snapshots, or user files; encrypt data for coercion; sabotage systems; or otherwise harm a user's computer or data.
No miners, fork bombs, resource exhaustion, or workflows that disable firewalls, backups, updates, logs, monitoring, or security services.
No account takeover, brute force, exploit chaining, vulnerability weaponization, or extraction from third-party systems without authorization.
No denial-of-service, botnet, spam, abusive traffic, scraping designed to bypass access controls, or infrastructure misuse.
No phishing, fake login capture, deceptive payment flows, identity theft, marketplace manipulation, or other transactional deception.
No voter intimidation, election-worker harassment, false voting-process information, election-office impersonation, or deceptive AI campaign impersonations.
No harmful manipulation, exploitation of vulnerabilities, social scoring, criminal-risk prediction, facial-recognition database scraping, protected-characteristic biometric categorization, or workplace/education emotion recognition.
No generic marketplace skills that automate, rank, score, approve, deny, or recommend consequential outcomes for employment, housing, credit, insurance, education, public benefits, healthcare triage, immigration, or law enforcement.
No generic marketplace skills that act as a doctor, lawyer, investment adviser, or other licensed professional by giving personalized diagnoses, prescriptions, legal advice or representation, investment recommendations, or auto-trading directives.
No skills that help evade OFAC, sanctions, export controls, embargoes, restricted-party screening, KYC, or AML checks.
No money laundering, terrorist financing, mule accounts, transaction structuring, crypto-mixer abuse, or concealment of illegal proceeds.
No forged IDs, passports, visas, licenses, bank statements, pay stubs, invoices, diplomas, prescriptions, or altered documents used to deceive.
No trafficking, sale, procurement, shipping, or distribution of illegal drugs, weapons, stolen goods, counterfeit goods, or unlawful services.
No skills that help make, convert, procure, deploy, or use firearms, silencers, machine-gun conversion devices, destructive devices, explosives, select agents, toxins, or chemical, biological, or radiological weapons.
No paywall, DRM, copy-protection, license-check, or copyright-management circumvention; no piracy, cracked software, or unauthorized copying or redistribution of protected works.
No child sexual abuse material, sexual exploitation, sextortion, non-consensual intimate imagery, AI nudification/digital forgeries, or related abuse workflows.
No non-consensual personal-data harvesting, people-finder datasets, covert tracking, deanonymization, or surveillance of private individuals through contact, location, biometric, device, or identity data.
No doxxing, stalking, swatting, non-consensual personal-data collection, or instructions that target private individuals.
No skills built to bypass Sealed containment, evade moderation, or provide operational instructions for physical harm.
02 Listing review path
The seller must certify the skill is lawful, authorized, non-harmful, and non-infringing before upload.
The server screens the sealed submission before it can list. Rejections return policy codes only; the skill body is never echoed in the response or audit trail.
Seller-submitted skills stay out of the public catalog and cannot run until a reviewer completes the fixed safety/legal checklist and applies a current policy approval stamp.
Durable rows without a current screening stamp fail closed. They must be re-screened and, when required, reviewed before traffic.
03 Enforcement
We may refuse publication, remove a listing, suspend access, hold payouts while investigating, or terminate an account when a skill violates this policy, the Terms, or the law. Severe or repeated abuse may be referred to the appropriate platform, payment, enterprise, or legal channel.
Sanctions, export-control, AML, KYC, and restricted-party controls apply across buyers, sellers, skills, and payouts. We may block account access, seller onboarding, skill publication, runs, wallet activity, or Connect cash-out when required for sanctions/export compliance, payment-network rules, Stripe requirements, law-enforcement/legal obligations, or marketplace safety.
Good-faith defensive, educational, compliance, and authorized internal security work can be allowed when it is scoped to lawful systems and does not provide operational abuse instructions.