docs(research): add Agent Bench Lab evaluation handoff

[t3chn]

Summary

• Add Agent Bench Lab fit, evaluation handoff, and blocker fields to Research Radar experiment proposal templates.
• Clarify that Radar output remains intake-only and cannot trigger Agent Bench Lab repo changes without approval.
• Document how code-intel-kernel R&D hypotheses should map to Agent Bench Lab suites, scorers, run validity, and compare metrics.

Issue

• N/A

Test plan

• python3 -m json.tool research-radar/templates/digest.json >/dev/null
• git diff --check
• scripts/run-deterministic-tests.sh

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 70f5b018d5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Ensure the collector emits the new handoff fields

When run_daily.py promotes any item with score >= 85, build_experiment_candidate() still returns only the old keys (title, source_url, source_type, hypothesis, minimal_reversible_change, evaluation_plan, stop_condition, and reason_not_to_implement_immediately). With these new required template fields, automated daily JSON/Markdown reports for that scenario will violate the updated Research Radar contract and omit the Agent Bench Lab fit/blocker that the guardrails now require, so the generator should be updated alongside the template.

Useful? React with 👍 / 👎.