feat(runner): add local command runner MVP#23
Merged
Conversation
Why: - Agent Bench Lab needs a product-neutral way to run real external agent setups against existing task families, not only score prebuilt sample artifacts. - The runner must preserve the benchmark visibility boundary by separating agent-visible task packets from scorer-visible fixtures. What changed: - Add `agent-bench run` and a local command runner that creates `task_packet/`, `artifacts/`, `run.json`, `trace.jsonl`, and `score.json`. - Exclude scorer-only/private-looking files from task packets, including `check_config.json`, while scoring against the original fixture and produced artifacts. - Add a mock smoke agent, runner tests, `make run-smoke`, CI wiring, schema event/status updates, and docs. Testing: - `make validate` - `make test` - `make smoke` - `make compare-smoke` - `make if01-smoke` - `make data01-smoke` - `make doc01-smoke` - `make sup01-smoke` - `make api01-smoke` - `make lifecycle-check` - `make mutation-smoke` - `make hardening-check` - `make run-smoke` - `make leak-check` - `python3 -m ruff check .` - `git diff --check` - tracked-file audit for private/generated/sensitive paths Risk: - moderate - command execution is intentionally local and generic; future provider/browser/MCP adapters should not bypass the task-packet visibility boundary. Related: #22
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
agent-bench runas a product-neutral command adapter for external local agent setups.task_packet/,artifacts/,run.json,trace.jsonl, andscore.jsonper run.check_config.json, answer keys, hidden labels, canaries, private scorer configs, and expected values out of the task packet.make run-smoke, CI wiring, runner tests, and local runner docs.Non-goals
Test plan
make validatemake testmake smokemake compare-smokemake if01-smokemake data01-smokemake doc01-smokemake sup01-smokemake api01-smokemake lifecycle-checkmake mutation-smokemake hardening-checkmake run-smokemake leak-checkpython3 -m ruff check .git diff --checkReview focus
src/agent_bench_lab/runner.py: task packet visibility boundary and command execution flow.tests/test_runner.py: timeout, missing artifacts, redaction, CLI, and scorer-only exclusion coverage.docs/21-local-agent-runner.md: ensure this is framed as a command adapter, not a provider/runtime layer.Risks
task_packet != scorer fixtureboundary.Breaking changes
Follow-ups
v0.7.1or becomes part of a later milestone.Merge strategy recommendation