Add adversarial safety fixtures for service runtime (shared helpers services)

[haasonsaas]

Summary

Exercise prompt/tool/data poisoning and fail-closed behavior for the repo's most sensitive agent-facing path.

This issue was generated from an org-wide EvalOps mining pass on 2026-05-10 07:57 UTC. It combines live GitHub repo signals with a per-repo arXiv search. Treat the research links as grounding for a concrete implementation, not as a request for a literature review.

Repo Evidence

• Repository description: Shared Go runtime helpers for EvalOps services
• Tree signals: 0 docs files, 4 workflows, 1 proto files, 11 test-like files.
• README.md:47 includes latent-spec language: The automation intentionally stays conservative and only advances the patch line. If maintainers need to start a new minor or major line, cut that seed tag manually first and the workflow will continue from there on later merges.
• README.md:91 includes latent-spec language: Use this package directly when a service needs retry behavior but still wants to own the actual bootstrap logic and logging. This is the pattern used by
• README.md:94 includes latent-spec language: services that want retry logs around each failed database attempt, but still need one place to register shutdown work such as tracing flushes.
• README.md:203 includes latent-spec language: Use OpenAndInit when a service needs to run bootstrap logic after the DB is reachable, such as:
• README.md:298 includes latent-spec language: Helpers for the shared EvalOps mTLS contract.
• README.md:308 includes latent-spec language: Use this package when a service needs the same client/server TLS file-path contract that memory, registry, meter, and audit share today.

Research Grounding

Repo axes: infra, governance, security, evaluation

Search keywords: ctx, err, startup, helpers, when, bootstrap, use, cfg, client, context, mtls, testutil

• arXiv:2604.03262v1 AI Governance Control Stack for Operational Stability: Achieving Hardened Governance in AI Systems (Horatio Morgan), 2026.
• arXiv:2503.15577v1 Navigating MLOps: Insights into Maturity, Lifecycle, Tools, and Careers (Jasper Stone, Raj Patel, Farbod Ghiasi, Sudip Mittal, Shahram Rahimi), 2025.
• arXiv:2601.20415v1 An Empirical Evaluation of Modern MLOps Frameworks (Jon Marcos-Mercadé, Unai Lopez-Novoa, Mikel Egaña Aranguren), 2026.
• arXiv:2604.24801v2 Architectural Observability Collapse in Transformers (Thomas Carmichael), 2026.
• arXiv:2502.15859v4 AI Governance InternationaL Evaluation Index (AGILE Index) 2024 (Yi Zeng, Enmeng Lu, Xin Guan, Cunqing Huangfu, Zizhe Ruan, Ammar Younas), 2025.
• arXiv:2510.21203v1 The Nuclear Analogy in AI Governance Research (Sophia Hatz), 2025.
• arXiv:2305.14865v1 A Game-Theoretic Framework for AI Governance (Na Zhang, Kun Yue, Chao Fang), 2023.
• arXiv:2407.01557v1 AI Governance and Accountability: An Analysis of Anthropic's Claude (Aman Priyanshu, Yash Maurya, Zuofei Hong), 2024.
• arXiv:2512.11541v1 A Multi-Criteria Automated MLOps Pipeline for Cost-Effective Cloud-Based Classifier Retraining in Response to Data Distribution Shifts (Emmanuel K. Katalay, David O. Dimandja, Jordan F. Masakuna), 2025.
• arXiv:1902.03439v1 PoliFi: Airtime Policy Enforcement for WiFi (Toke Høiland-Jørgensen, Per Hurtig, Anna Brunstrom), 2019.

What To Build

• Add adversarial fixtures for deployment drift, credentials, and privileged workflow inputs.
• Document the intended fail-closed behavior and any allowed degraded-mode fallback.
• Add regression coverage that proves unsafe inputs do not silently reach the privileged path.

Acceptance Criteria

• A short design note names the repo-specific workflow, threat or correctness model, and the research assumptions being adopted.
• A runnable check, fixture, or verifier exercises the new contract in CI or an equivalent local command documented in the repo.
• The implementation emits or stores enough evidence for a downstream agent/operator to cite inputs, decisions, and outputs.
• At least one negative/degraded-mode case is covered so failures are observable rather than silently accepted.
• Documentation links the new behavior to the relevant EvalOps platform primitive or explicitly records why this repo remains standalone.

Notes

• Generated issue 3/5 for evalops/service-runtime by evalops_org_miner.py.
• Before implementation, confirm the sampled latent-spec snippets still match main; this issue intentionally cites exact file paths/lines where the mining pass saw them.