AgentPROV corpora: Apache-2.0 studies (DTap, AgentOps-Bench and AgentDojo-format studies)

Execution-provenance corpora built by AgentPROV v0.1.0 from run logs published under the Apache License 2.0 (DTap-Bench trajectories, AgentOps-Bench, and AgentDojo-format studies): each execution as an AgentPROV Execution Record, semantic annotations and W3C PROV bundles, validated and hash-chained. This deposit draft holds 6 AgentPROV corpus version(s), prepared offline by agentprov package --deposit for the release date 2026-10-30. Nothing has been uploaded or published. Corpus agentops-bench 0.1.0: AgentOps-Bench v1.0 pilot: 8 models x 100 tasks x 3 conditions x 3 runs (Zenodo 19875693), in-scope part. 3360 execution records, of which 101 are outcome-only records. Licence class: Apache-2.0. Content digest sha256:fbacf3cef7187e8c1796772895caafc445da6c016e0e7552e932e5684d446d7a; root digest sha256:0ce0dee36f7dd925ea2efa7ae57f467aad89c47734020818a543d2bbdce16515. Corpus dtap-bench 0.1.0: DTap-Bench agent trajectories (AI-Secure) with the DecodingTrust-Agent-Platform task configurations. 67708 execution records, of which 24 are outcome-only records. Licence class: Apache-2.0. Content digest sha256:b12d51a958e3365e1661a4e2195f3bb7a3ca6bbd45f0433ee95fe72fed15c2c0; root digest sha256:2fdc9bcbc5281d48338756936f98b38b522f22021da65d79aebf30969d8c9ec4. Corpus geryon-agentdojo 0.1.0: Geryon: AgentDojo (Progent fork) v0.k11 logs of conditions A and B (luiacuaniello/geryon). 3179 execution records, of which 0 are outcome-only records. Licence class: Apache-2.0. Content digest sha256:0d46a69b8784588841c92572abb784662f0d71e0f970f2a3399a84b9135eb9eb; root digest sha256:e11b366645fe5c07ac5bdc43c7b2eea947a9aa3ffab99187d7aa37511b2ca214. Corpus interbolt-agentdojo 0.1.0: Interbolt on AgentDojo: AgentDojo 0.1.35 logs of a provenance tool-call gate (deconvolute-labs/interbolt-agentdojo). 2016 execution records, of which 0 are outcome-only records. Licence class: Apache-2.0. Content digest sha256:8e3e643f87b2cce03eabc2b7d9653e85293c3dcd4659e2de63cc735198d32450; root digest sha256:28fb58d49ca2fa6b72b472c65d48795dd2bcc5a925f5bb1a90cd1957564e578e. Corpus tacit-agentdojo 0.1.0: TACIT AgentDojo trajectories: AgentDojo 0.1.35 logs of two local models (CSSLab/Tacit). 2070 execution records, of which 0 are outcome-only records. Licence class: Apache-2.0. Content digest sha256:594add0c5fb4674ed90fb4d7de7cd9f3e3967fdf758932a1cf04b9e9b60bad84; root digest sha256:c0c6121c602baf830cbb35a16eab4d25dc161bf0be7c5ac93461a85d6d08759f. Corpus yockcheng-prompt-injection-harness 0.1.0: Prompt injection harness: AgentDojo 0.1.35 workspace logs of four Claude models (yockcheng/prompt-injection-harness). 280 execution records, of which 0 are outcome-only records. Licence class: Apache-2.0. Content digest sha256:171b568c6cd8f633fd1bbc877b47d02233ebb994a950ae27555d16f30eb1b79f; root digest sha256:c94a0ac8b5a99a412b10068a182b98f98e4aadd0b1675f0d45252c72645a7911. Counts are computed from each corpus manifest at sealing. Each DATASHEET states what the records do and do not establish; a statement about the data is scoped as it is there. To verify: compare each archive's sha256 or md5 with DEPOSIT-MANIFEST.json, run agentprov verify on the archives (archive mode, no extraction; H.5.5), and compare each root digest with the one published outside the deposit (the corpus tag annotation and the paper).