Colby McHenry 9cf671acf5 chore(agent-eval): add per-repo WITH-only why-Read benchmark harness ...

bench-why-repo.sh runs one README repo, WITH-codegraph only, N runs, each
appending a why-Read diagnostic ("if you Read/Grep, explain why codegraph fell
short; end with a ## Why I read section"). Honors EFFORT (default high) and
kills the per-repo daemon before each run so it serves the current dist. The
WITHOUT baseline is codegraph-independent and lives in the README, so it's not
re-run. Used to validate the explore reshape: all 7 README repos, 28/28 runs,
0 Read / 0 Grep on effort=high.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-02 08:32:09 -05:00

..

add-lang

feat: add Lua and Luau language support (#273)

2026-05-21 09:28:00 -05:00

agent-eval

chore(agent-eval): add per-repo WITH-only why-Read benchmark harness

2026-06-02 08:32:09 -05:00

build-bundle.sh

fix: prevent V8 turboshaft WASM Zone OOM during indexing (#298, #293) (#322)

2026-05-22 05:39:24 -05:00

extract-release-notes.mjs

chore(release): refine release-notes extractor (#181)

2026-05-19 12:05:59 -05:00

local-install.sh

feat(installer): multi-target — Claude Code, Cursor, Codex CLI, opencode (#162)

2026-05-17 19:26:09 -05:00

npm-sdk.js

feat(npm): restore programmatic/embedded SDK API (#354) (#603)

2026-05-31 19:05:40 -05:00

npm-shim.js

fix: self-heal missing platform bundle from GitHub Releases (#303) (#335)

2026-05-22 11:38:28 -05:00

pack-npm.sh

feat(npm): restore programmatic/embedded SDK API (#354) (#603)

2026-05-31 19:05:40 -05:00

prepare-release.mjs

docs(claude): rewrite release section for auto-promote workflow + link-ref on promote (#437)

2026-05-26 03:25:06 -05:00