{"id":"ia.tool-definition-security","name":"Tool definition security","verify":{"command":null,"rescan":"isagentready results {domain} --checkpoint ia.tool-definition-security","expect":null},"fix":{"steps":[{"action":"Add readOnlyHint and untrustedContentHint annotations to each tool.","order":1,"language":"javascript","code_example":"annotations: { readOnlyHint: true, untrustedContentHint: true }"},{"action":"Remove tokens, internal URLs and personal data from tool names and descriptions.","order":2,"language":null,"code_example":null},{"action":"Pass an AbortSignal so that the page can unregister tools.","order":3,"language":null,"code_example":null}],"summary":"Add trust annotations to WebMCP tools and keep secrets out of tool definitions.","skill":{"name":"security-trust","anchor":"ia.tool-definition-security"}},"category":"security","aggregate":null,"references":[{"title":"WebMCP W3C Proposal","kind":"guide","url":"https://webmachinelearning.github.io/webmcp/"},{"title":"OWASP Sensitive Data Exposure","kind":"guide","url":"https://owasp.org/www-project-web-security-testing-guide/"}],"stage":"act","lifecycle":{"introduced":"2026-08-24","revised":[],"retired":null,"replaced_by":null},"methodology_version":"2026-10-22","evidence_tier":"experimental","evidence_source":{"title":"WebMCP W3C Proposal","url":"https://webmachinelearning.github.io/webmcp/"},"why":"Tool definitions are visible to AI agents and may be cached or logged. Explicit trust metadata, bounded origin exposure, and lifecycle cancellation help agents invoke tools safely.","profiles":["api_tool_readiness","agentic_commerce"],"score_role":"informational","legacy_id":"5.10","i18n":{"nl":{"name":"Veiligheid van tooldefinities","steps":["Voeg de annotaties readOnlyHint en untrustedContentHint toe aan elke tool.","Verwijder tokens, interne URL's en persoonsgegevens uit toolnamen en beschrijvingen.","Geef een AbortSignal mee, zodat de pagina tools kan afmelden."],"why":"Tooldefinities zijn zichtbaar voor AI-agents en kunnen worden opgeslagen of gelogd. Expliciete vertrouwensmetadata, beperkte origin-blootstelling en annulering helpen agents om tools veilig te gebruiken.","fix_summary":"Voeg vertrouwensannotaties toe aan WebMCP-tools en houd geheimen uit tooldefinities.","verify_expect":null}},"ar_practice_ids":[],"strength":"MAY","ar_mapping":"none","evidence_type":"recommendation","detail_url":"https://isagentready.com/en/methodology/checkpoints/ia.tool-definition-security","maturity":"emerging","applicability":{"mode":"detected_surface","condition":null,"surfaces":["auth"]}}