{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:UBETCRBNMAWWFN4BIOYBYFTMJK","short_pith_number":"pith:UBETCRBN","schema_version":"1.0","canonical_sha256":"a04931442d602d62b78143b01c166c4aa6e3a83b517293f79f58836dc4e603b7","source":{"kind":"arxiv","id":"2505.05849","version":4},"attestation_state":"computed","paper":{"title":"AgentVigil: Generic Black-Box Red-teaming for Indirect Prompt Injection against LLM Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Chenguang Wang, Dawn Song, Tianneng Shi, Vincent Siu, Wenbo Guo, Xuandong Zhao, Yuzhou Nie, Zhe Ye, Zhun Wang","submitted_at":"2025-05-09T07:40:17Z","abstract_excerpt":"The strong planning and reasoning capabilities of Large Language Models (LLMs) have fostered the development of agent-based systems capable of leveraging external tools and interacting with increasingly complex environments. However, these powerful features also introduce a critical security risk: indirect prompt injection, a sophisticated attack vector that compromises the core of these agents, the LLM, by manipulating contextual information rather than direct user prompts. In this work, we propose a generic black-box fuzzing framework, AgentVigil, designed to automatically discover and explo"},"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":"2505.05849","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2025-05-09T07:40:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"33f27a157a9a854263e92eb6ebf78c3ad3753a1aa94e3ae71deb1bae222fa91a","abstract_canon_sha256":"ce009bd67a70ed26e0dc6c37b1430f0aaf5f995ac16fe2c83864d2e54db8ea51"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:21:46.271541Z","signature_b64":"NsIr/eYGxTR1I2u/Yd/JeCqIZPuATZ2kcifd7+Or19XjSr/WX7f25IX/XWjaPHJ8mfWZ7pArE8QJdEPtq3IXCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a04931442d602d62b78143b01c166c4aa6e3a83b517293f79f58836dc4e603b7","last_reissued_at":"2026-07-05T11:21:46.270807Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:21:46.270807Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AgentVigil: Generic Black-Box Red-teaming for Indirect Prompt Injection against LLM Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Chenguang Wang, Dawn Song, Tianneng Shi, Vincent Siu, Wenbo Guo, Xuandong Zhao, Yuzhou Nie, Zhe Ye, Zhun Wang","submitted_at":"2025-05-09T07:40:17Z","abstract_excerpt":"The strong planning and reasoning capabilities of Large Language Models (LLMs) have fostered the development of agent-based systems capable of leveraging external tools and interacting with increasingly complex environments. However, these powerful features also introduce a critical security risk: indirect prompt injection, a sophisticated attack vector that compromises the core of these agents, the LLM, by manipulating contextual information rather than direct user prompts. In this work, we propose a generic black-box fuzzing framework, AgentVigil, designed to automatically discover and explo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.05849","kind":"arxiv","version":4},"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/2505.05849/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":"2505.05849","created_at":"2026-07-05T11:21:46.270867+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.05849v4","created_at":"2026-07-05T11:21:46.270867+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.05849","created_at":"2026-07-05T11:21:46.270867+00:00"},{"alias_kind":"pith_short_12","alias_value":"UBETCRBNMAWW","created_at":"2026-07-05T11:21:46.270867+00:00"},{"alias_kind":"pith_short_16","alias_value":"UBETCRBNMAWWFN4B","created_at":"2026-07-05T11:21:46.270867+00:00"},{"alias_kind":"pith_short_8","alias_value":"UBETCRBN","created_at":"2026-07-05T11:21:46.270867+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22864","citing_title":"When AUC 0.998 Is Not Enough: A Candidate Evaluation Protocol for Hidden-State Probes of Indirect Prompt Injection in Multimodal Computer-Use Agents","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12737","citing_title":"PI-Hunter: Automated Red-Teaming for Exposing and Localizing Prompt Injections","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10525","citing_title":"Assessing Automated Prompt Injection Attacks in Agentic Environments","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.14027","citing_title":"Same-Origin Policy for Agentic Browsers","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00566","citing_title":"Same Payload, Different Channel: Measuring Trust Asymmetry in Tool-Using Language Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2504.11703","citing_title":"Progent: Securing AI Agents with Privilege Control","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19192","citing_title":"Hallucination as Exploit: Evidence-Carrying Multimodal Agents","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18991","citing_title":"Agent Security is a Systems Problem","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19192","citing_title":"Hallucination as Exploit: Evidence-Carrying Multimodal Agents","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18991","citing_title":"Agent Security is a Systems Problem","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06393","citing_title":"Constraining Host-Level Abuse in Self-Hosted Computer-Use Agents via TEE-Backed Isolation","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK","json":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK.json","graph_json":"https://pith.science/api/pith-number/UBETCRBNMAWWFN4BIOYBYFTMJK/graph.json","events_json":"https://pith.science/api/pith-number/UBETCRBNMAWWFN4BIOYBYFTMJK/events.json","paper":"https://pith.science/paper/UBETCRBN"},"agent_actions":{"view_html":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK","download_json":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK.json","view_paper":"https://pith.science/paper/UBETCRBN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.05849&json=true","fetch_graph":"https://pith.science/api/pith-number/UBETCRBNMAWWFN4BIOYBYFTMJK/graph.json","fetch_events":"https://pith.science/api/pith-number/UBETCRBNMAWWFN4BIOYBYFTMJK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK/action/storage_attestation","attest_author":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK/action/author_attestation","sign_citation":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK/action/citation_signature","submit_replication":"https://pith.science/pith/UBETCRBNMAWWFN4BIOYBYFTMJK/action/replication_record"}},"created_at":"2026-07-05T11:21:46.270867+00:00","updated_at":"2026-07-05T11:21:46.270867+00:00"}