Skip to content

Add canary string to public benchmark tasks and datasets#2

Open
jonmaaku-rev wants to merge 6 commits into
mainfrom
feat/canary-string
Open

Add canary string to public benchmark tasks and datasets#2
jonmaaku-rev wants to merge 6 commits into
mainfrom
feat/canary-string

Conversation

@jonmaaku-rev

Copy link
Copy Markdown
Collaborator

Summary

  • Add a permanent canary GUID as a single source of truth (scripts/canary.py, documented in CANARY.md).
  • Stamp the canary into all task files and the synthetic data in artifacts/data.zip (+ CANARY.md sidecar).
  • Add an idempotent stamper (scripts/add_canary.py) and a canary check in scripts/validate_repo.py that blocks un-stamped tasks.
  • Document the policy in CONTRIBUTING.md.

Follows the BIG-bench canary standard (as adopted by Terminal-Bench / Harbor). The GUID is permanent — do not change it.

Related issue

https://app.devrev.ai/devrev/works/ISS-348138

Test plan

  • make validate passes (ruff + structure + canary check + digest match).
  • python scripts/add_canary.py --check reports 0 to stamp (fully idempotent).
  • data.zip SHA-256 matches dataset.toml.

Add the permanent canary GUID as a single source of truth in scripts/canary.py (documented in CANARY.md) and an idempotent stamper script with a --check dry-run. Follows the BIG-bench canary standard.
Add the canary to every public task file (instruction, README, task.toml, Dockerfile, criteria, trajectory, test.sh), using the correct embedding per file type. Mechanical change; no task logic altered.
Stamp the synthetic data in artifacts/data.zip and add a CANARY.md sidecar; regenerate the archive and dataset.toml digests to match.
Fail validation when a required task file, CANARY.md, or the data.zip sidecar is missing the canary, and verify dataset.toml digests. Lands after stamping so validation stays green.
Document the canary requirement (enforced by make validate) in CONTRIBUTING.md and reference it in README.md.

@roycabezas roycabezas left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants