Rewrite PR comments to sound human, not corporate-AI

Rewrites or generates code review comments so they read like a teammate wrote them, not a politeness engine - direct on the issue, concrete on the fix, kind on the human. Triggers on /unslop-review, /review, "review this PR," "humanize review," "de-slop this comment," or automatically when reviewing pull requests.

The default comment shape is L<line>: <severity prefix> <observation>. <fix>. (multi-file: <file>:L<line>: ...; a line range: L88-140: ...), with optional severity prefixes used when severity matters: bug: for code that's broken or will break, risk: for something that works today but is fragile (performance, race conditions, missing tests), nit: for style/naming/dead-code/while-you're-here items, and q: for a genuine question rather than a hidden complaint - e.g. a real example given in the skill's source: "L42: bug: findUser returns undefined when no match. Guard before user.email or early-return 404."

It drops throat-clearing ("I noticed that...", "It seems like..."), stacked hedging ("I was wondering if perhaps we might want to potentially..."), polite-padding ("I would kindly suggest...", "just a small suggestion..."), per-comment praise ("Nice work on this function but..."), restating the diff instead of making a point, and bare opinions with no fix attached ("This is bad" with no suggestion). It keeps exact line numbers/ranges, identifiers in backticks, a concrete fix or question, and a "why" only when the fix isn't self-evident. Tone is human, not corporate - calibrated uncertainty ("I think", "probably") is fine, performative softening is not.

Security findings (CVE-class, auth, secrets), architecture disagreements needing real discussion, onboarding context for a new contributor, and cases where the honest answer is "this is fine" get a short paragraph of full prose instead of a one-liner, then the rest of the review resumes terse. If a change is solid with nothing concrete to flag, the entire comment is just LGTM on its own line, no boilerplate around it.

Use it whenever code review comments need writing or an existing corporate-toned draft needs de-slopping into something a teammate would actually say. It only produces comment text - no commits, no git push, no auto-approve, no linter runs - and severity must stay honest (never downgrade a bug to a nit just to soften the message).

Input is a diff/PR to review or an existing draft comment to rewrite. Output is paste-ready review comments - one per line, or a clearly separated list - each following the L<line>: <severity> <observation>. <fix>. shape, or a bare LGTM when there's nothing to flag.

Format-only tooling around standard PR review workflows (line-referenced comments with severity prefixes) - no direct integration with a specific git host or CI system beyond producing comment text ready to paste into one.

Reviewers who want direct, specific, human-sounding PR feedback (location, issue, concrete fix) without the corporate hedging and empty praise that make automated or template-driven review comments feel evasive.