Publish Standard60 task snapshot

[bingran-you]

Summary

• publish the exact 60-task Standard60 snapshot under tasks/
• keep Discord as an env-only fixture under example_tasks/
• enforce native task.md + oracle/ + verifier/ layout across all 68 public task packages
• bump the public base image to 0.2.0 so auth, Discord, and Stripe CLIs are included
• add maintainer, release, public-data, and task-snapshot CI gates

Why

The public repo exposed only 12 of the current 60 canonical tasks. The published
0.1.0/latest image also predated mock-auth, mock-discord, and
mock-stripe, so those task packages could not build against the remote image.

Public-data review

• replaced contributor-style emails and handles with deterministic
  skillsbench.test identities
• replaced assignable-looking SSNs with impossible 000-00-... fixtures
• sanitized captured Slack test-workspace emails to example.test
• removed credential-bearing database URLs
• added a repo-wide CI guard for real key formats, assignable SSNs, local paths,
  private-repo links, and sensitive files
• retained only documented localhost mock credentials, Stripe test cards, and
  the explicitly non-secret deterministic mock-auth signing key

Validation

• python3 -m unittest tests/test_standard60_tasks.py tests/test_public_data_hygiene.py
• 68/68 BenchFlow structural task checks passed
• 121 targeted evaluator regression tests passed
• auth/Stripe oracle regression: 16/16 passed against locally built 0.2.0
• CI builds 0.2.0, smokes example images, and runs one oracle task for each of
  the eight public mock environments

Release impact

After merge, dispatch Publish Base Image, verify all eight mock-* CLIs,
publish GHCR 0.2.0 and latest, then tag and release v0.2.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 39fda0fa21

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Fail verification when any service binary is missing

In the publish workflow, this inner sh script doesn't enable errexit; inside the loop, a failing command -v is overwritten by later iterations. If a pushed image is missing any binary except the last one (mock-stripe), the “Verify remote pull” step can still exit with the final successful check and pass while the published base image is incomplete. Add set -e or explicitly exit 1 on a missing command before the loop continues.

Useful? React with 👍 / 👎.