← ClaudeAtlas

review-dispatch-body-afterlisted

Fresh-context dispatch shapes after the reviewer slot, inlined by the review verbs. Not for direct invocation.
m4ttstack/skills · ★ 0 · Code & Development · score 75
Install: claude install-skill m4ttstack/skills
**Nothing under Reviewer above** (the slot is unbound): apply `superpowers:requesting-code-review`. Fill its `code-reviewer.md` placeholders -- `DESCRIPTION` (the caller's description), `PLAN_OR_REQUIREMENTS` (the requirements), `BASE_SHA` / `HEAD_SHA` (the commit range ends) -- then append the standard blocks. If the superpowers plugin is not installed, say so and stop. **Content under Reviewer above**: follow it, still appending the standard blocks. Either branch dispatches its template as written: placeholders filled, standard blocks appended, every other section untouched. What to look hardest at is the fresh reviewer's call; a hunch about where this change is weak travels as a setup observation, not a rewritten checklist. Both branches return the same shape, stated in the dispatched prompt: **Strengths / Critical / Important / Minor / Assessment**. Callers assemble drafts out of those buckets, so a bound provider owes them too. ## Adjudicator shape Fill `${CLAUDE_SKILL_DIR}/references/adjudicator.md` (`{DIFF_RANGE}`, `{REQUIREMENTS}`, `{THREADS}`), append the standard blocks, and dispatch ONE fresh-context `general-purpose` subagent with it. One dispatch judges all threads together. ## Red flags | Thought | Reality | |---|---| | "It's tiny, just eyeball it" | Size is not the variable; the authoring context is. | | "We're in a hurry" | Building the dispatch is one message; the inline review is the slow path when it misses. | | "I wrote it this session, so I know th