perf(mentoring): trim author skill routing metadata

[anbv29]

Summary

Shortened the description and invocation guidance for good-first-issue-author by removing details already covered in the skill body. The existing trigger phrases, scope restrictions, and confirmation requirements are preserved.

The always-on estimate drops from 273 to 168 tokens, below the 200-token budget. The measured full-file count drops from 3,609 to 3,503 tokens. The skill body is unchanged.

Updated the mentoring plugin’s published token estimate from ~0.5k to ~0.4k.

Validation

• Target skill validation passes.
• Token stamps and the mentoring family estimate check pass.
• git diff --check passes.
• Verified that the skill body, protected frontmatter, and quoted trigger phrases are unchanged.
• Before and after eval runs each returned 13 CLI authentication errors; no cases were graded.
• Whole-tree prek and lychee checks remain pending because hook initialization stalled on Windows.

Scope

This is the first pass for one skill in the mentoring family. The remaining three skills and any body wording changes are outside this PR.

Refs #1351

[Kaap10]

Thanks @anbv29!

Verified the diff at code level:

• Frontmatter for good-first-issue-author drops to ~173 always-on tokens (cleanly within the 200 budget) while keeping all three quoted routing triggers, maintainer confirmation gate, and safety rules intact.
• The skill body is completely untouched and measured_tokens is updated to 3503.
• docs/setup/marketplace.md table estimate is synced to ~0.4k.
• All CI checks pass.

LGTM!

[anbv29]

Hey @Kaap10
Is it ready for the merge or any other changes to make? I'll look it right away if needed.

[onlyarnav]

lgtm