Add adversarial safety fixtures for orbit agent (brutally honest high orbit startup)

[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: A brutally honest "high‑orbit" startup advisor you can text or run from the CLI. Built with DSPy, it provides opinionated, YC-style advice and financial tools for founders.
• Tree signals: 0 docs files, 1 workflows, 0 proto files, 8 test-like files.
• README.md:15 includes latent-spec language: - 🧠 Best-of-N + Rerank: Generate multiple drafts and pick the best via a critic. - 🧪 Evals & Rubrics: Personas, rubrics, overlap penalty, and CSV/MD summaries.
• README.md:66 includes latent-spec language: - models list [--provider openai|anthropic]: List available model IDs. - eval run --dataset <yaml> --out <jsonl>: Run evals and save results. - eval report <jsonl>: Show overall summary.
• README.md:67 includes latent-spec language: - eval run --dataset <yaml> --out <jsonl>: Run evals and save results. - eval report <jsonl>: Show overall summary. - eval grade --dataset <yaml> --results-path <jsonl> --out <jsonl>: Rubric grading.
• README.md:68 includes latent-spec language: - eval report <jsonl>: Show overall summary. - eval grade --dataset <yaml> --results-path <jsonl> --out <jsonl>: Rubric grading. - eval summary --input-path <jsonl> [--csv-out <csv>] [--md-out <md>]: Export summaries.
• README.md:69 includes latent-spec language: - eval grade --dataset <yaml> --results-path <jsonl> --out <jsonl>: Rubric grading. - eval summary --input-path <jsonl> [--csv-out <csv>] [--md-out <md>]: Export summaries.
• README.md:140 includes latent-spec language: ## Evals & Self‑Grading

Research Grounding

Repo axes: infra, governance, security, evaluation

Search keywords: jsonl, cli, run, evals, eval, str, orbit_agent, export, list, yaml, orbit, personas

• arXiv:2604.04749v1 AI Trust OS -- A Continuous Governance Framework for Autonomous AI Observability and Zero-Trust Compliance in Enterprise Environments (Eranga Bandara, Asanga Gunaratna, Ross Gore, Abdul Rahman, Ravi Mukkamala, Sachin Shetty), 2026.
• arXiv:2604.26152v1 AI Observability for Large Language Model Systems: A Multi-Layer Analysis of Monitoring Approaches from Confidence Calibration to Infrastructure Tracing (Twinkll Sisodia), 2026.
• arXiv:2604.17092v1 AI Observability for Developer Productivity Tools: Bridging Cost Awareness and Code Quality (Happy Bhati, Twinkll Sisodia), 2026.
• arXiv:2604.03262v1 AI Governance Control Stack for Operational Stability: Achieving Hardened Governance in AI Systems (Horatio Morgan), 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:2503.15577v1 Navigating MLOps: Insights into Maturity, Lifecycle, Tools, and Careers (Jasper Stone, Raj Patel, Farbod Ghiasi, Sudip Mittal, Shahram Rahimi), 2025.
• arXiv:2407.01557v1 AI Governance and Accountability: An Analysis of Anthropic's Claude (Aman Priyanshu, Yash Maurya, Zuofei Hong), 2024.
• arXiv:2510.21203v1 The Nuclear Analogy in AI Governance Research (Sophia Hatz), 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.

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/orbit-agent 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.