ValentinNikolaev/agent-skills / agent-plugins/skills/discussion-reply
discussion-reply
VerifiedPrepare grounded replies to social-media or forum discussions, from source context through exact-copy review. Not private correspondence, code-hosting replies, original posts, or wording-only edits.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.1 KB |
| agents/openai.yaml | config | 253 B |
| references/contribution-value.md | markdown | 4.6 KB |
| references/review-state.md | markdown | 4.0 KB |
references/contribution-value.md
View on GitHub# Contribution value check
Use this for an analytical reply: a finding, objection, correction, proposed fix, technical answer, or claim about a consequence. It starts where the source-fit and level checks end. A reply can match the topic, be accurate, concrete, and well written, and still give its reader nothing to use. This check adds no reviewer, rubric, weight, threshold, or repair; it runs inside the exact-copy assessment the project already requires.
## Scope
- Support, acknowledgment, humor, and a short human reaction are not gated by it. Judge the recognition the text actually offers under the owning rubric. Do not ask for advice, a technical point, or an action.
- When both are present, check the part that carries the reply. Warmth around a weak claim does not exempt it. A technical aside does not turn a complete recognition into analysis.
- No kind of observation is excluded in advance, documentation and naming findings included. A small finding can pass when the reply shows why it matters here. Its subject is never the reason for the verdict.
## Independent pass before ratings
Before assigning any rating, each independent reviewer answers all five points from the exact copy and the source packet, without other reviewers' answers or earlier totals. Without a required panel, one local reviewer does the same and does not claim independence. Quote the reply's own words in every answer. An explanation that exists only in review notes does not count.
1. **Question and link.** What task or question is the author working on? What question does the exact reply answer? Is the link between them substantive? Report the result of the project's fit gates here instead of repeating them.
2. **Reader gain.** What does the addressee or reader receive: a basis for a decision, understanding of a mechanism or limit, a significant correction, a well-founded objection, or another gain? Quote the words that deliver it.
3. **Trace.** For a finding or proposed fix, follow observation, consequence, then significance for the discussed task. Mark each step confirmed (in the reply or the sources) or assumed. Do not supply a serious effect to justify a small inconsistency. A consequence found only in a reviewer's explanation is assumed.
4. **Effect of acting.** If the fix were made, what changes for the author's decision, and what remains of their main problem? Is the result large enough to deserve this branch's attention?
5. **Marginal contribution.** What does the text add to the source and to the strongest visible comments? Verifiability, a novel detail, a primary source, technical terms, and smooth style do not establish usefulness.
Give **pass** or **revise** with the deciding evidence. Revise when the link to the author's task is not substantive, the reader's gain cannot be quoted from the reply, the decisive consequence is assumed, acting would leave the author's decision and main problem unchanged at a scale that does not justify the branch's attention, or the text adds nothing beyond the source and visible comments.
## Effect on ratings
A revise verdict is a failed gate, not a deduction. Apply the project's existing failed-gate limit on its conversational-value criterion if it has one; otherwise treat the reply as not accepted. A high total, reviewer agreement, a favorable detector result, or flawless prose does not override it. A pass admits the reply to ratings. It raises no rating and satisfies no fit, level, or voice gate. Weights, thresholds, floors, reviewer count, and the repair budget stay as the project set them.
## Peer review and final selection
In the anonymized review, each reviewer tests the strongest argument that the text is valuable and the strongest objection to it, using the reply and the sources. The coordinator resolves any substantive value objection from the reply's wording and the sources before adopting ratings. Votes, averages, and agreement do not resolve it: one reviewer's unresolved revise verdict stands.
At final selection the coordinator repeats the check on the exact copy to be delivered, in its own words, as if the user had asked what the practical point of the reply is. Record the quoted basis for a pass. A substantive edit reopens the check for that copy.
## Repair or withdraw
A repair supplies what the copy lacks: a shown consequence, the decision it informs, or a different contribution. Polishing does not. It counts within the existing repair budget and returns the changed copy to exact-copy assessment. If no candidate proves its value, withdraw the idea, leave the slot empty or the reply undelivered, and report the reason. Do not polish a weak idea to a passing score, invent a consequence, or add an evaluator.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |