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/review-state.md
View on GitHub# Version and review controls
Use one ephemeral working record rather than repeatedly rebuilding the brief. Retain it across a long-task checkpoint when needed, excluding private source dumps and council transcripts. `editorial-writing` shares these controls; for a standalone piece, read the recipient and placement fields as its reader and destination.
Record:
- source packet identity/revision, complete relevant content or accessible evidence references, coverage and freshness limits;
- speaker, recipient, placement, contribution, voice constraints, required translations and deliverable count;
- exact draft text or an unambiguous version/hash for each option and translation;
- required gates with `pending`, `passed`, `failed`, or `unavailable`, the checked version, and evidence;
- baseline draft, substantive repair count, remaining budget, and the user's specific objection with its wording, reader effect and resolution.
A hash identifies a version; it does not prove a check ran. Record actual assessment or service output separately, naming the reviewer or service and the version it evaluated. A later stage may reuse the source packet only after confirming every reviewer still has the same complete content. Restore missing context before judging; a summary alone is not equivalent to a previously required full source.
## Invalidate only the affected work
| Change | Revisit |
| --- | --- |
| New source clarification, changed recipient, purpose, or interpretation | Source/intent decision and affected drafts and assessments |
| Changed reasoning, stance, structure, or substantive wording | Local edit checks and required exact-copy assessment for that option |
| Typo or meaning-preserving punctuation | Coordinator comparison; preserve unchanged editorial judgments when policy permits |
| Any change to text submitted to a required exact-text service | Service evidence for the new submitted text, even if editorial judgments remain valid |
| Another option changes | That option; keep unrelated valid assessments |
Before ratings, each reviewer describes the action they read in the exact text and what the recipient gets from it, then compares that reading with the supplied brief. Include the agreed brief and all required evidence from the outset. Hold back the editor’s defence of the wording.
When independent reviewers are required, record their individual findings before sharing anonymized responses for peer challenge. Reuse the initial reviewers for revised versions within the host budget. Returning reviewers are not blind new readers; disclose their prior exposure, keep earlier scores and any coordinator-preferred verdict out of fresh packets, and resolve material objections from the current copy and evidence. Agreement is not factual verification or proof of the author's preference.
## Bound repairs
Use the project's finite revision budget. Otherwise permit at most two substantive repair batches after the first complete draft, then report unresolved issues. Before a batch, identify the unresolved objection, the observable correction, and whether the required evidence exists. Stop and report a blocker if missing evidence or unchanged constraints prevent progress. Do not reset the baseline by moving a file, renaming the draft, or calling a substantive rewrite polish. Never reroll unchanged copy for a higher score, change the rubric, or invent evidence to obtain a pass.
If repeated structural repairs leave the same reader objection, reconsider the contribution before spending a remaining iteration. A changed contribution returns to the host's intent stage. A better score or a different service label alone cannot close the user's objection. Report an unresolved goal separately from successful delivery or saving.
Keep interpretation, editorial quality, service observations, and any requested engagement or future-post assessment separate. An undefined rubric yields an unresolved numeric assessment, not an invented number. Follow the owner's explicit qualitative fallback when available.
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 |