terminal-bench: stop uploading benchmark artifacts - #4002
Merged
Conversation
Incident remediation: the terminal-bench-results-* artifacts published by this workflow contained live provider API keys in plaintext, and artifacts on a public repo are downloadable by any authenticated GitHub user. Remove the "Upload benchmark results" step (jobs/ + benchmark.log) and the now-unused "Set artifact name" step. Run output stays on the runner; the BigQuery result uploads are unchanged. This is the immediate stop-the-bleeding measure only. The underlying cause (secrets on the process argv via Harbor's `env NAME=value` serialisation) is untouched and needs a follow-up, and previously retained artifacts must be deleted separately. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Aug 28, 2026
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The terminal-bench workflow uploads the full
jobs/tree (includingjobs/*/agent/command-0/stdout.txt) as a 30-day artifact on a public repo. Agent command lines — which currently contain forwarded provider credentials in the process argv — therefore land in a downloadable artifact (31 of 64 retained artifacts contain provider keys).This PR removes the two steps responsible: the
Upload benchmark resultsartifact upload (jobs/ + benchmark.log) and the now-unusedSet artifact namestep. Nightly runs keep working; they just no longer publish anything.Local tooling note:
download_run_logs.pyandtbench_utils.pystop finding new runs once this merges — they read from workflow artifacts. (Real fix for argv exposure — no secrets on argv, per-lane forwarding only, mask at capture — is tracked separately.)