Colby McHenry 77845c747d test(agent-eval): show every run's residual and tool mix, not just the median (CG-7) ...

A median over 2-3 runs hides swings big enough to flip a repo's sign. On vscode
the without-arm ranged 40k to 67k and the with-arm 59k to 65k across two runs;
the deciding variable is the tool mix, since a with-arm run that reads files ON
TOP of calling explore pays for both.

2026-08-04 14:52:07 -05:00

..

ab-adoption.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

ab-hook.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

ab-impl.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

ab-new-vs-baseline.sh

feat(explore): dynamic-dispatch boundary surfacing — announce where a flow ends instead of guessing edges (#687) (#835)

2026-06-12 11:26:33 -05:00

ab-sufficiency.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

arms-F.sh

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

arms-matrix.sh

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

audit.sh

feat(eval): add agent-eval harness and /audit + /publish Claude skills

2026-05-20 09:39:17 -05:00

bench-readme.sh

test(agent-eval): measure residual context occupancy, over multi-turn sessions (CG-7)

2026-08-04 14:35:56 -05:00

bench-why-repo.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

block-read-hook.sh

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

hook-settings.json

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

itrun.sh

feat(resolution): gin middleware-chain synthesizer + Opus 4.8 benchmark refresh (#547)

2026-05-28 23:41:11 -05:00

offload-eval-3arm.sh

test(agent-eval): add output-style A/B harness, cost/token analyzer, and DISALLOW/REP_START controls

2026-06-19 16:43:10 -05:00

offload-eval-cost.mjs

test(agent-eval): add output-style A/B harness, cost/token analyzer, and DISALLOW/REP_START controls

2026-06-19 16:43:10 -05:00

offload-eval-effort.mjs

feat(mcp): pare default tool surface to codegraph_explore alone + redux-thunk synthesizer

2026-06-19 02:15:14 -05:00

offload-eval-frontload-matrix.sh

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-frontload.sh

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-ground-truth.json

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-hook.mjs

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-judge.mjs

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-matrix.sh

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-metrics.mjs

fix(mcp): bold labels instead of ATX headings in tool results (#778) (#951)

2026-06-22 12:06:35 -05:00

offload-eval-refs1.sh

feat(mcp): pare default tool surface to codegraph_explore alone + redux-thunk synthesizer

2026-06-19 02:15:14 -05:00

offload-eval-setup.sh

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval-styles.sh

test(agent-eval): add output-style A/B harness, cost/token analyzer, and DISALLOW/REP_START controls

2026-06-19 16:43:10 -05:00

offload-eval-summarize.mjs

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

offload-eval.md

test(agent-eval): persist offload accuracy/adoption eval harness + front-load hook

2026-06-18 22:30:30 -05:00

parse-arms.mjs

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

parse-bench-readme.mjs

test(agent-eval): show every run's residual and tool mix, not just the median (CG-7)

2026-08-04 14:52:07 -05:00

parse-run.mjs

test(agent-eval): self-test the occupancy math, and fix ratio calibration under shedding (CG-7)

2026-08-04 14:41:13 -05:00

parse-session.mjs

feat(eval): add agent-eval harness and /audit + /publish Claude skills

2026-05-20 09:39:17 -05:00

probe-context.mjs

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

probe-explore.mjs

fix(mcp): bold labels instead of ATX headings in tool results (#778) (#951)

2026-06-22 12:06:35 -05:00

probe-node.mjs

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

probe-sweep.mjs

feat(mcp): multi-module Go trace-quality + small-repo retrieval tuning (#494)

2026-05-28 12:38:03 -05:00

probe-trace.mjs

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00

redirect-read-hook.sh

feat(mcp): codegraph_node reads files like the Read tool — offset/limit, byte-parity (#738)

2026-06-08 13:48:42 -04:00

repro-concurrent-explore.mjs

feat(mcp): off-load read-tool dispatch to a worker pool to fix concurrent-call timeouts (#1002)

2026-06-26 15:42:18 -05:00

repro-daemon-clients.mjs

feat(mcp): off-load read-tool dispatch to a worker pool to fix concurrent-call timeouts (#1002)

2026-06-26 15:42:18 -05:00

run-agent.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

run-all.sh

test(agent-eval): measure residual context occupancy, over multi-turn sessions (CG-7)

2026-08-04 14:35:56 -05:00

run-arms.sh

chore(agent-eval): standing A/B model policy — sonnet + high effort, never Opus/Fable (#816)

2026-06-11 19:39:16 -05:00

seq-matrix.mjs

Release 0.9.4: framework-aware routing + dynamic-dispatch coverage + retrieval improvements (#365)

2026-05-24 04:41:04 -05:00