kud/claude-plugins/discuss
Claude Code and agent skill by kud/claude-plugins
- community
Checking that the SKILL.md file is still in the repository.
Reads the annotations you captured in revu (from .revu.json, via a live headless export) and talks the review through as a senior engineer — leading with blockers, respecting triage state — then offers to clean up any export file.
Install
npx vskill@latest install kud/claude-plugins/discussSKILL.md
You are discussing a code review the user captured with revu. Talk it through — you do not implement here (that is the implement skill). Lead with the most severe items and honour the user's triage.
Step 1 — Get the review data
Prefer the live, structured export — it carries severity, status, source, and the captured code range, and never needs a file:
Run that in the current directory (add --against <branch> if the user is reviewing a PR branch). It prints JSON: { prompt, comments: [{ file, startLine, endLine, side, severity, status, source, text, code }] }.
Fallbacks, in order, if that command is unavailable or errors:
- Read
.revu.jsondirectly (revu autosaves it on every annotation — same fields, minuscode). - Read
revu-review.md(the markdown export; parse the annotated sections).
If none exist, tell the user:
"No revu review found. Run revu in this repo, annotate the diff (press ↵ on a line), and I'll pick it up — no export needed."
Then stop.
Step 2 — Read the review prompt
If a prompt is present (JSON prompt field, or the text before the --- in revu-review.md), treat it as the user's intent for the review and let it frame your responses.
Step 3 — Order and filter by triage
- Skip annotations whose
statusisdismissedorresolvedunless the user explicitly asks to include them — the user has already triaged those away. - Order the rest by severity:
blocker→concern→nitpick→ unset. Discuss blockers first. - Note the
source: an annotation withsource: "agent"came from another AI pass — treat it as a proposal the human is triaging, not a settled decision.
Step 4 — Discuss each annotation
Work through the surviving annotations in severity order. For each:
- Name it by
fileand line(s), and its severity if set (e.g. "blocker — src/auth.ts:42"). - Respond as a senior reviewer: surface concrete risks, suggest specific improvements, and ask a clarifying question where the intent is unclear.
- Reference the actual
codein the snippet — be precise, not generic.
Treat the user's annotations as the only agenda. Do not produce a general review of the whole diff; engage only with what they marked.
Step 5 — Offer to clean up
If you read revu-review.md, ask at the end: "Delete revu-review.md?" — and if yes, remove it with trash revu-review.md (fallback rm). Never delete .revu.json: it is the live autosave and must persist across sessions.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingplugins/revu/skills/discuss/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/kud/claude-plugins/discuss)