docs: fix benchmark repository references

[ntudy]

Summary

• update benchmark documentation to use the bundled FrontierAgent paths
• remove stale internal-repository and Git-subtree references
• align FrontierSearchBench scorer attribution and FrontierChallenge site documentation with the public repository

Validation

• checked all affected local Markdown links and image targets
• verified benchmark registry entries and documented task/query counts
• ran Python syntax checks and git diff --check
• completed a defect-first review with no findings

Test notes

• full pytest suite was not run because the local project virtual environment is not configured

[Copilot]

Pull request overview

Updates benchmark documentation and scorer attribution to reflect bundled FrontierSearchBench and FrontierChallenge paths.

Changes:

• Removes stale standalone-repository and Git-subtree references.
• Updates benchmark paths, ownership notes, and site deployment guidance.
• Points scorer attribution headers to the FrontierAgent repository.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
README.md	Describes bundled benchmarks.
docs/framework.md	Updates architecture tree.
docs/eval-frontier-search.md	Revises benchmark ownership and workflow references.
benchmarks/README.md	Updates benchmark descriptions and registry terminology.
benchmarks/public/runner/score_frontier_search.py	Updates scorer comment terminology.
benchmarks/public/README.md	Documents bundled sibling benchmark.
benchmarks/frontierchallenge/site/README.md	Corrects site build and deployment instructions.
benchmarks/frontier_search_bench/UPSTREAM.md	Replaces obsolete subtree provenance.
benchmarks/frontier_search_bench/README.md	Documents in-repository paths and setup.
benchmarks/frontier_search_bench/eval/verifiable/scorers/query_38/auto_scorer.py	Updates HTTP attribution.
benchmarks/frontier_search_bench/eval/verifiable/scorers/query_15/auto_scorer.py	Updates HTTP attribution.
benchmarks/frontier_search_bench/eval/verifiable/README.md	Updates directory terminology; credential location remains inaccurate.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[zhanghanduo]

LGTM