LN: DeBellis et al. (2026) — Authenticated Workflows: A Systems Approach to Protecting Agentic AI
Bibliographic Reference
Citation: DeBellis, M., Rivera, C., & Shankar, K. (2026). Authenticated workflows: A systems approach to protecting agentic AI. arXiv:2602.10465. https://arxiv.org/abs/2602.10465 (This paper is in the PUMA project knowledge PDFs as “Authenticated_Workflows_A_Systems_Approach_to_Protecting_Agentic_AI_2602.10465v1.pdf”)
Pass 1 — Bird’s Eye View (5 Cs)
| C | Assessment |
|---|---|
| Category | System proposal + security framework |
| Context | Addresses authentication and authorisation in multi-step agentic workflows |
| Correctness | Systems-level analysis. Proof-of-concept implementation. |
| Contributions | (1) Authentication model for agentic workflows; (2) Preventing unauthorised agent actions; (3) Audit trail for agentic decision chains |
| Clarity | Good. |
Relevance: ⭐⭐⭐
PUMA’s governance layer (Constitution Article 4: HITL; Article 5: open source) needs authentication mechanisms. Relevant for Stage 5 Smart PMO where agents interact with real Jira/GitHub APIs.
PUMA Connection
When PUMA Stage 5 agents interact with production Jira instances, authenticated workflows are mandatory. This paper’s framework informs PUMA’s governance spec for the Smart PMO. Reference for Section 1.3 (ethics) and architecture spec.