scripts/check_kernel_freshness: include kernel_dir in stale report items

[Dev-X25874]

Closes #879

What does this PR do?

Adds kernel_dir to each line in the Slack freshness report so stale alerts identify which local directory needs updating, not just the upstream URL.

Motivation

Several kernels share the same upstream URL. For example flash-attn2, flash-attn3, flash-attn4, layer-norm, rotary, and vllm-flash-attn3 all point to the same flash-attention repo. When any go stale, the current report prints the same URL multiple times with no indication of which directory is behind, making the alert unactionable. The kernel_dir key is already populated in every result dict — it was just never used when building the output string.

Changes

• scripts/check_kernel_freshness.py: one-line format string change in _format_freshness_report to prepend kernel_dir to each report line
• scripts/test_kernel_freshness_lookup.py: added test_freshness_report_includes_kernel_dir — calls _format_freshness_report with two results sharing a source URL, asserts both kernel_dir values appear in the output

Testing

cd scripts
python test_kernel_freshness_lookup.py

New test fails on original code, passes after the fix.

Checklist

• This PR is linked to an issue that was discussed and approved
• I have tested these changes locally
• New/changed functionality has test coverage
• LLM disclosure:
  • I did not use an LLM to create this PR.
  • I used an LLM for assistance while creating this PR.
  • This PR was mostly or completely generated by an LLM.

[github-actions]

Hi @Dev-X25874, thanks for your interest in contributing!

This project requires that pull request authors are vouched, and you are not in the list of vouched users.

This PR will be closed automatically. See https://github.com/huggingface/kernels-community/blob/main/CONTRIBUTING.md for more details.

[Dev-X25874]

Hey @sayakpaul, just wanted to flag that the security-audit check failure doesn't seem related to the changes in this PR.

Looking at the job logs, the claude -p --model claude-opus-4-7 command on line 143 exits with code 1 before even completing the analysis — the "Notify Slack" step is skipped entirely, which suggests the Claude CLI itself crashed rather than producing findings. Likely an API key issue or transient infra failure on the runner.

Would you be able to re-run the security audit job or confirm it's a known infra issue? Happy to address any actual findings if they come through. Thanks!