Fix log instructions for claude

[fafk]

Description

Log format has changed so I updated instructions how to use it for Claude.

[gemini-code-assist]

Code Review

This pull request updates documentation for log queries to reflect changes in the query format. I have one suggestion to improve placeholder consistency in docs/COW_ORDER_DEBUG_SKILL.md for better clarity.

_{Note: Security Review has been skipped due to the limited scope of the PR.}

[jmg-duarte]

PR looks good, does raise some questions in my head

I wonder if you do the following it would improve "search performance" (not speed-wise, result quality)

1. Provide a link to the VL query language spec (something like https://docs.victoriametrics.com/victorialogs/querying/)
2. Provide meaning mappings for labels
3. Some examples (like you do already)

I also wonder if, instead of providing a large skills/CLAUDE.md, we provide a "tree" of files that it can get and the scenarios it should reach out for it, it would reduce the context size and thus avoid compactions

[fafk]

PR looks good, does raise some questions in my head

I wonder if you do the following it would improve "search performance" (not speed-wise, result quality)

1. Provide a link to the VL query language spec (something like docs.victoriametrics.com/victorialogs/querying)
2. Provide meaning mappings for labels
3. Some examples (like you do already)

I also wonder if, instead of providing a large skills/CLAUDE.md, we provide a "tree" of files that it can get and the scenarios it should reach out for it, it would reduce the context size and thus avoid compactions

I think it might. So far with the "all" field it's been performing well. Claude is also very smart so it uses tail to cut the size and if there is a lot of output it writes it to a temp file and runs a python script over it to get insights. One needs to experiment with these things to see what works.