Publish agent handoff provenance study

[chenmingtang830]

What changed

• adds the public Agent-Native Research Artifact study package for Proofpress agent handoffs
• publishes the technical report, study card, release manifest, and explicit licensing boundary
• includes aggregate and selected machine-readable evidence, the current result figure, and SHA-256 checksums
• records the bounded mechanism finding and separates it from general product-efficacy claims

Why

This makes the completed local evaluation reviewable from the production Proofpress repository and gives future citations a stable, auditable package.

Validation

• ARA Seal Level 1: 86/86 checks passed through the recorded Proofpress carrier-comment compiler adapter
• 21 portable Markdown artifacts inspected
• 95 JSON files and 6 YAML files parsed
• 0 missing relative Markdown links
• non-capsule secret scan clean
• included Python files compiled
• 125 package-file SHA-256 checks verified
• git diff --check passed

Release boundary

This is an early, bounded mechanism evaluation, not a peer-reviewed general-efficacy claim. The fixed release tag agent-handoff-artifact-provenance-rc1 should be created from the final main commit after this PR is merged.

[github-actions]

Proofpress

studies/agent-handoff-artifact-provenance/README.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/RELEASE_MANIFEST.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/STUDY_CARD.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/STUDY_LICENSE.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/PAPER.md
74cc7326 → dc633f61 ▍ 123 added 7 removed

• 🟢 new · --- title: "Proofpress: An Artifact Provenance Protocol · (文首)
• 🟢 new · - "Readable handoffs reduce rediscovery, but do not esta · (文首)
• 🟢 new · abstract: "Prior research on agent memory studies how sy · (文首)
• 🟢 new · # Proofpress: An Artifact Provenance Protocol for Agent
• 🟢 new · **Working paper — Chenming Tang, Independent Researcher. · Proofpress: An Artifact Provenance Proto
• 🟢 new · > Takeaway: A handoff tells a successor agent what i · Proofpress: An Artifact Provenance Proto
• 🟢 new · ## Abstract
• 🟢 new · Prior research on agent memory studies how systems retai · Abstract
• 🟢 new · We introduce **Proofpress, an artifact provenance protoc · Abstract
• 🟢 new · We evaluated Proofpress using document-handoff tasks ada · Abstract
• 🟢 new · These results support a bounded conclusion: readable han · Abstract
• 🟢 new · ## 1. Introduction
• 🟢 new · Agent memory and handoff research has largely focused on · 1. Introduction
• 🟢 new · Agent handoff is already an explicit systems pattern rat · 1. Introduction
• 🟢 new · Provenance research addresses a complementary problem. [ · 1. Introduction
• 🟢 new · The risk created by mutable artifacts is likewise establ · 1. Introduction
• 🟢 new · This paper introduces Proofpress, an artifact proven · 1. Introduction
• 🟢 new · 1. Continuity: missing context forces a successor to · 1. Introduction
• 🟢 new · A strong memo can reduce the first problem. The second r · 1. Introduction
• 🟢 new · Proofpress records accepted artifact versions, claims, a · 1. Introduction
• …and 110 more

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/UPSTREAM.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/evidence/README.md
30dce380 → 0b1c5103 ▍ 6 modified

• 🔵 mod · Release evidence map
• 🔵 mod · Release evidence map
• 🔵 mod · Release evidence map
• 🔵 mod · Release evidence map
• 🔵 mod · Source and design inputs
• 🔵 mod · Source and design inputs

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/evidence/tables/derived_main_results.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/claims.md
fcdb716c → 43487f93 ▍ 14 added

• 🟢 new · ## C14: The contribution is a joint agent-handoff evalua
• 🟢 new · - Statement: Prior work already covers obsolete memo · C14: The contribution is a joint agent-h
• 🟢 new · ## C13: Perfect separation on the primary fixture is mec
• 🟢 new · - Statement: Because Proofpress natively verifies ar · C13: Perfect separation on the primary f
• 🟢 new · ## C11: Native artifact binding has confirmatory support
• 🟢 new · Statement and proof: In the preregistered OpenRouter · C11: Native artifact binding has confirm
• 🟢 new · ## C12: The bounded stale-binding result replicated acro
• 🟢 new · Statement and proof: In a frozen-protocol formal fol · C12: The bounded stale-binding result re
• 🟢 new · ## C09: The tested native binding mechanism replicated o
• 🟢 new · - Statement: In the three selected public HiEviDR `a · C09: The tested native binding mechanism
• 🟢 new · ## C10: R2 semantic faults show recovery-path, not demon
• 🟢 new · - Statement: In the selected HiEviDR-derived semanti · C10: R2 semantic faults show recovery-pa
• 🟢 new · ## C08: Native artifact binding can block stale-artifact
• 🟢 new · - Statement: In the frozen APEX-derived SPA stale-bi · C08: Native artifact binding can block s

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/concepts.md
6bb4cb45 → cdd9bef5 ▍ 8 added

• 🟢 new · ## K08: Proofpress
• 🟢 new · A deterministic artifact-provenance and admission layer. · K08: Proofpress
• 🟢 new · ## K09: Design-aligned mechanism fixture
• 🟢 new · An evaluation fixture whose injected fault directly inst · K09: Design-aligned mechanism fixture
• 🟢 new · ## K10: Identity is not currentness
• 🟢 new · A digest, signature, or content-addressed identifier can · K10: Identity is not currentness
• 🟢 new · ## K11: Proofpress artifact provenance protocol
• 🟢 new · - Notation: `Proofpress = (record, link, transfer, c · K11: Proofpress artifact provenance prot

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/experiments.md
6a87adfe → aa2013c5 ▍ 43 added 3 modified 4 removed

• 🟢 new · ## E15: Controlled unchanged-versus-outdated artifact ad
• 🟢 new · - Verifies: C11, C13. - Evidence: `evidence/tabl · E15: Controlled unchanged-versus-outdate
• 🟢 new · ## E16: Receiver-model robustness on the frozen admissio
• 🟢 new · - Verifies: C12, C13. - Evidence: `evidence/cros · E16: Receiver-model robustness on the fr
• 🟢 new · ## E17: Visible-contradiction and selective-recovery bou
• 🟢 new · - Verifies: C09, C10, C13. - Evidence: `evidence · E17: Visible-contradiction and selective
• 🟢 new · ## E18: Handoff continuity without artifact change
• 🟢 new · - Verifies: C02, C13. - Evidence: `evidence/know · E18: Handoff continuity without artifact
• 🟢 new · ## E19: Structured related-work scoping review
• 🟢 new · - Verifies: C14. - Evidence: `logic/related_work · E19: Structured related-work scoping rev
• 🟢 new · ## Comprehensive local experiment synthesis (2026-08-12)
• 🟢 new · docs/LOCAL_EXPERIMENT_RESULTS.md now synthesizes the c · Comprehensive local experiment synthesis
• 🟢 new · The comprehensive write-up keeps all valid outcomes in t · Comprehensive local experiment synthesis
• 🟢 new · ## Registered next program: knowledge handoff debt under
• 🟢 new · Status: early testing completed on 2026-08-12; forma · Registered next program: knowledge hando
• 🟢 new · ### Knowledge handoff early-testing results (2026-08-12)
• 🟢 new · Track 1: 9/9 DeepSeek responses were valid across th _· Knowledge handoff early-testing results _
• 🟢 new · Interpretation: the early evidence supports two sepa _· Knowledge handoff early-testing results _
• 🟢 new · ## Interpretation audit and next validity gate (2026-08-
• 🟢 new · The primary APEX-derived experiment is a design-aligned · Interpretation audit and next validity g
• …and 30 more

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/problem.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/related_work.md
264600dd → 48631d47 ▍ 23 added

• 🟢 new · ## RW01: Handoff Debt
• 🟢 new · - Type: baseline and boundary. - Source: https:/ · RW01: Handoff Debt
• 🟢 new · ## RW02: AgentAsk
• 🟢 new · - Type: failure taxonomy. - Source: https://acla · RW02: AgentAsk
• 🟢 new · ## RW03: W3C PROV
• 🟢 new · - Type: imported provenance model. - Source: htt · RW03: W3C PROV
• 🟢 new · ## RW04: Software consume-time verification
• 🟢 new · - Type: mechanism precedent. - Source: in-toto, · RW04: Software consume-time verification
• 🟢 new · ## RW05: Portable and provenance-aware agent memory
• 🟢 new · - Type: closest agent-system precedent. - Source · RW05: Portable and provenance-aware agen
• 🟢 new · ## Review status and method
• 🟢 new · This is a structured primary-source scoping review curre · Review status and method
• 🟢 new · ## Review synthesis: the broad ideas are established
• 🟢 new · Prior work already evaluates obsolete memories and evolv · Review synthesis: the broad ideas are es
• 🟢 new · ## Closest agent-memory and handoff work
• 🟢 new · | Work | What it evaluates | Relationship and remaining · Closest agent-memory and handoff work
• 🟢 new · ## Provenance, versioning, and consume-time admission
• 🟢 new · | Work or standard | Established mechanism | Boundary re · Provenance, versioning, and consume-time
• 🟢 new · ## Benchmark validity and interpretation
• 🟢 new · | Work | Guidance applied here | |---|---| | [Measuring · Benchmark validity and interpretation
• …and 3 more

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/solution/algorithm.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/solution/architecture.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/solution/constraints.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/logic/solution/heuristics.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/src/artifacts.md
bc123264 → 204b54bd ▍ 2 modified

• 🔵 mod · Experiment artifact index Δ 1→2
• 🔵 mod · Experiment artifact index

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/src/configs/model.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/src/configs/training.md
fewer than two ledger versions — nothing to diff yet
⚪ no claims recorded for this version (snapshot ran without --claims)

studies/agent-handoff-artifact-provenance/ara/src/environment.md
7bce8de2 → c5a27562 ▍ 1 modified

• 🔵 mod · Environment

changelog: publish agent handoff provenance study
⚪ no claims recorded for this version (snapshot ran without --claims)