recipe-maintainers/agent-orchestrator-benchmark
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: add builder-adversary-lean variant; runner takes variant args

Browse Source 
- bump engine submodule to e0425e6 (adds builder-adversary-lean: context hygiene
  + enforced per-gate review)
- run-harness-bench.sh: accept variant names as CLI args to run a subset
This commit is contained in:
mfowler
2026-06-14 21:43:11 +00:00
parent 0fa3d726a5
commit cca5c895b2
2 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
engine

Submodule engine updated: 985d33dd51...e0425e6108

3
run-harness-bench.sh
View File

@ -20,7 +20,8 @@ RESULTS="$BENCH_DIR/RESULTS-harness.md"
TIMEOUT="${BENCH_TIMEOUT:-3000}" # seconds per variant
POLL=60
GIT_ID=(-c user.email=bench@example.com -c user.name=bench)
VARIANTS=(builder-adversary builder-adversary-min builder-adversary-stateless)
VARIANTS=(builder-adversary builder-adversary-min builder-adversary-stateless builder-adversary-lean)
[ $# -gt 0 ] && VARIANTS=("$@") # run only the variants named on the command line, if any
# pre-trust a work dir in ~/.claude.json so interactive claude (in tmux) skips the workspace-trust
# dialog (--dangerously-skip-permissions only skips it for redirected/headless output). Atomic merge: