code-review evals: re-measure the comment track and cover the standards slot in scanner prompts

[grixu]

Follow-up from #8 review.

• The comment track (evals/prompts/review.txt, eval-0 … eval-4) still uses the old opener and was not re-measured under @anthropic-ai/claude-agent-sdk 0.3.252. It likely needs the same treatment as quality.txt: name the skill explicitly (code-review:comment-review) and ask for a report-only run.
• The three scanner-track prompts (security.txt, performance.txt, spec.txt) hardcode the standards slot to "none", so the new lenses' <standards> path — any lens raising a standards finding — has no eval coverage. Add a standards-slot variant to at least one scanner-track eval.

https://claude.ai/code/session_01BcNztjBNBtjwfTRzbYEZjR

[pullfrog]

  ^{｜ Build this ➔ ｜ Make a plan ➔}