Repository navigation
[1/3] fix(benchmarks): track docs pages agents read across all harnesses - #5330
Merged
Merged
Conversation
Contributor
benpankow
force-pushed
the
benchmarks/track-docs-reads
branch
from
October 26, 2026 20:25
ddf6bb8 to
39e11a2
Compare
benpankow
marked this pull request as ready for review
October 26, 2026 21:33
benpankow
force-pushed
the
benchmarks/track-docs-reads
branch
from
October 26, 2026 21:41
b543207 to
0d848c5
Compare
Signed-off-by: benpankow <ben.pankow@vercel.com>
Signed-off-by: benpankow <ben.pankow@vercel.com>
Signed-off-by: benpankow <ben.pankow@vercel.com>
Signed-off-by: benpankow <ben.pankow@vercel.com>
benpankow
force-pushed
the
benchmarks/track-docs-reads
branch
from
October 26, 2026 22:00
0d848c5 to
fb5fc8d
Compare
prha
approved these changes
Oct 26, 2026
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
We want to know which shipped docs pages help coding agents in the authoring benchmarks, and which pages they never open.
scripts/analyze.mjsalready reported docs reads, but it parsed an older raw transcript shape. With the native Claude Code, Codex, and OpenCode harnesses, it finds zero docs reads. The script now reads tool calls from agent-eval's harness-neutraltranscript.json.--docsranks pages by how many runs opened them and lists shipped pages that no run opened.These numbers cover benchmark tasks only, not real usage. The published results file does not change.
Validation
Ran the old and new
analyze.mjs --docson a fixture with one native-shaped transcript per harness. Each transcript readsREADME.md. Old: no pages. New:README.mdin 3/3 runs, plus 108 unread shipped pages.Added
scripts/analyze.test.mjs. It buildstranscript.jsonwith agent-eval's ownparseTranscriptfor each harness and checks the aggregate and the unread list.pnpm testinapps/benchmarks: 38 tests pass.pnpm typecheckinapps/benchmarks, plusoxfmtandoxlinton the changed files: pass.Ran the benchmark on this PR with kimi-k3 (OpenCode), claude-sonnet-5 (Claude Code), and gpt-6-sol-high (Codex) on weather-tool and openapi-connection: 6/6 passed.
analyze.mjson the artifacts lists the pages each run opened. Every run entered throughREADME.md, then readtools/*orconnections/*for its case.README.md,tools/overview.mdxconnections/openapi.mdx,connections/overview.mdx,README.mdtools/human-in-the-loop.md×2,tools/overview.mdx×2,connections/overview.mdx,README.mdconnections/openapi.mdx,connections/overview.mdx,README.mdREADME.md,tools/human-in-the-loop.md,tools/overview.mdxconnections/openapi.mdx,connections/overview.mdx,README.mdTokens in include cache reads.
Checklist
evepackagegit commit --signoff)