Add PerspectiveGap to the research catalogue

[WhymustIhaveaname]

Summary

Adds PerspectiveGap as a research resource on information boundaries in multi-agent orchestration prompts, together with the required resource-quality scoresheet.

Contribution Type

• Resource, standard, framework, or research
• Tool or benchmark
• Case study
• Secure engineering pattern or architecture
• Governance, contribution, or repository maintenance
• Other

Evidence And Scope

• Source or evidence: https://arxiv.org/abs/2606.08878 and the public benchmark repository
• Producer or publisher: Sun et al., arXiv 2026
• Risks, behaviours, controls, or evaluation methods covered: role-specific information boundaries, omissions, cross-role leakage, and deterministic orchestration-prompt evaluation
• Maturity or evidence level: emerging academic benchmark with public scenarios, scoring code, results, and leaderboard
• Last-checked date, if external: 2026-07-25
• Limitations or caveats: evaluates prompt artefacts rather than downstream execution, tools, credentials, memory, or production outcomes

Review Focus

Please check the placement in the papers catalogue and the resource-quality scoring.

Checklist

• The change is relevant to agentic execution security.
• Claims are evidence-led and do not overstate repository maturity.
• British English is used except for established names such as AI Defense Plane.
• No unnecessary operational exploit detail is included.
• No copied source material is included when a summary and link would be enough.
• No generated outputs, built documentation, logs, traces, screenshots, reports, exports, or local artefacts are included.
• A corresponding scoresheet is included.

Validation completed with the repository's Remark lint, Markdown link check, and MkDocs build.