{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:CT43O3OJ442SKPL6QGYWWVZLTB","short_pith_number":"pith:CT43O3OJ","schema_version":"1.0","canonical_sha256":"14f9b76dc9e735253d7e81b16b572b987f27ee6b236b51e4e18991c4de47f7b2","source":{"kind":"arxiv","id":"2607.29254","version":1},"attestation_state":"computed","paper":{"title":"Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Jiayuxuan Yang, Minghui Pan, Yuanyuan Yuan, Yu Jiang, Zhenpeng Chen","submitted_at":"2026-07-31T10:25:04Z","abstract_excerpt":"AI agents extend large language models (LLMs) with external tools, enabling them to perform complex tasks and translate model outputs into consequential real-world actions. Yet LLMs often become substantially less safe when deployed as agents, and the source of this degradation remains poorly understood. In this paper, we identify schema-formatted tool specifications as a primary source of agent safety degradation and show, through white-box representation analysis, that they weaken the model's internal refusal signals and contribute to unsafe tool execution. Building on this finding, we propo"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.29254","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-31T10:25:04Z","cross_cats_sorted":[],"title_canon_sha256":"3b276f981f798eb39ddac5793d7a52a599af23008d8dfcbcc12274d5303f08af","abstract_canon_sha256":"46735e7f1cbb193f7a1f8dc827396db9eb8a0f6495a5f555407b8cce159e154e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-03T01:22:29.074653Z","signature_b64":"GBeS84iY690tTC16Lsg8aXULTPLmGmaTH+01TxOKMbMyvTcsT0pMtdvYQENBVZFqCUWZFiiCDma/6T1SiVaSAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"14f9b76dc9e735253d7e81b16b572b987f27ee6b236b51e4e18991c4de47f7b2","last_reissued_at":"2026-08-03T01:22:29.072953Z","signature_status":"signed_v1","first_computed_at":"2026-08-03T01:22:29.072953Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Jiayuxuan Yang, Minghui Pan, Yuanyuan Yuan, Yu Jiang, Zhenpeng Chen","submitted_at":"2026-07-31T10:25:04Z","abstract_excerpt":"AI agents extend large language models (LLMs) with external tools, enabling them to perform complex tasks and translate model outputs into consequential real-world actions. Yet LLMs often become substantially less safe when deployed as agents, and the source of this degradation remains poorly understood. In this paper, we identify schema-formatted tool specifications as a primary source of agent safety degradation and show, through white-box representation analysis, that they weaken the model's internal refusal signals and contribute to unsafe tool execution. Building on this finding, we propo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.29254","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.29254/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.29254","created_at":"2026-08-03T01:22:29.073857+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.29254v1","created_at":"2026-08-03T01:22:29.073857+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.29254","created_at":"2026-08-03T01:22:29.073857+00:00"},{"alias_kind":"pith_short_12","alias_value":"CT43O3OJ442S","created_at":"2026-08-03T01:22:29.073857+00:00"},{"alias_kind":"pith_short_16","alias_value":"CT43O3OJ442SKPL6","created_at":"2026-08-03T01:22:29.073857+00:00"},{"alias_kind":"pith_short_8","alias_value":"CT43O3OJ","created_at":"2026-08-03T01:22:29.073857+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB","json":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB.json","graph_json":"https://pith.science/api/pith-number/CT43O3OJ442SKPL6QGYWWVZLTB/graph.json","events_json":"https://pith.science/api/pith-number/CT43O3OJ442SKPL6QGYWWVZLTB/events.json","paper":"https://pith.science/paper/CT43O3OJ"},"agent_actions":{"view_html":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB","download_json":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB.json","view_paper":"https://pith.science/paper/CT43O3OJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.29254&json=true","fetch_graph":"https://pith.science/api/pith-number/CT43O3OJ442SKPL6QGYWWVZLTB/graph.json","fetch_events":"https://pith.science/api/pith-number/CT43O3OJ442SKPL6QGYWWVZLTB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB/action/storage_attestation","attest_author":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB/action/author_attestation","sign_citation":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB/action/citation_signature","submit_replication":"https://pith.science/pith/CT43O3OJ442SKPL6QGYWWVZLTB/action/replication_record"}},"created_at":"2026-08-03T01:22:29.073857+00:00","updated_at":"2026-08-03T01:22:29.073857+00:00"}