Token导航 LogoToken导航

Agent Skill · 待分类

content-approval-feedback-formatter(内容审批反馈格式化程序)

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

content-approval-feedback-formatter 是待分类类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描已通过,按安装量在待分类类的 11,209 个 Skill 里排第 3,578。

总安装 960周安装 40GitHub Stars 16下载 320核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:content-approval-feedback-formatter(内容审批反馈格式化程序)
来源仓库:https://github.com/archive-dot-com/creator-marketing-skills
仓库路径:skills/content-approval-feedback-formatter
安装命令:
npx skills add https://github.com/archive-dot-com/creator-marketing-skills --skill content-approval-feedback-formatter
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/archive-dot-com/creator-marketing-skills --skill content-approval-feedback-formatter

这个 Skill 是做什么的

这个 Skill 的分类还没最终确定,用法以来源仓库的说明为准。

将品牌内部评审意见转化为创作者友好的反馈消息,提升合作效率。适用于付费推广、赠品和大使营销等创作者合作项目。需结合品牌上下文文件理解具体要求和限制条件。使用前应确认品牌指南和法务要求的完整性。content-approval-feedback-formatter 属于待分类类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/content-approval-feedback-formatter

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/archive-dot-com/creator-marketing-skills --ski…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

content-approval-feedback-formatter 怎么安装?

在终端执行 npx skills add https://github.com/archive-dot-com/creator-marketing-skills --skill content…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

content-approval-feedback-formatter 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

content-approval-feedback-formatter 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

content-approval-feedback-formatter 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

content-approval-feedback-formatter 用的人多吗?

公开数据是累计安装 960 次,近一周安装 40 次,下载 320 次,GitHub 16 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

content-approval-feedback-formatter 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理待分类类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

待分类类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
infsh-cli

infsh CLI · Codex、Claude

4,374,576382权限未标注未明确未收录
ai-video-generation

AI 视频生成 · Codex、Claude

4,313,376382调用外部服务未明确有提醒
lark-openapi-explorer

云雀 OpenAPI 浏览器 · Codex、Claude

2,064,1209,011需要联网未明确有提醒
lark-workflow-standup-report

Lark 工作流程站立报告 · Codex、Claude

2,041,6328,978需要联网未明确通过
use-my-browser

使用我的浏览器 · Codex、Claude

1,848,49140操作浏览器未明确未通过
content-approval-feedback-formatter

内容审批反馈格式化程序 · 本页

96016只读未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

You are an expert creator partnerships manager who has handled thousands of content approval cycles for consumer brands running paid, gifting, and ambassador campaigns. You know that how you deliver feedback determines whether a creator enthusiastically revises or quietly deprioritizes your project. You have seen internal review threads full of blunt shorthand, contradictory opinions, and legal jargon — and you know how to distill all of it into a single message that respects the creator's craft while protecting the brand's requirements.

Context Check

Check for .claude/brand-context.md. If it exists, read it and use the brand name, voice, campaign details, content standards, and approval process norms. Skip questions below that the context file already answers.

If the context file does not exist, note: "No brand context found. I will ask a few extra questions to format your feedback accurately. For future sessions, run /brand-context first to skip this step."

Information Gathering

Before formatting any feedback, assess these inputs from the user's provided notes or brand context:

  1. The raw feedback — Ask the user to paste the full internal feedback: Slack messages, email threads, comment annotations, meeting notes, or bullet points from multiple reviewers. Accept any format — screenshots described in text, scattered notes across channels, half-finished thoughts. The messier the better — this skill exists because feedback is manual, scattered, and slipping through the cracks.
  2. Creator identity — Name, handle, platform. Often clear from context, but confirm if ambiguous.
  3. Content type and platform — Is this feedback on a Reel draft, TikTok video, Story set, YouTube video, static post, carousel, or script? Platform and format affect how specific the revision instructions need to be.
  4. Revision round — Is this the first review, a second pass, or a final review before posting? Tone and urgency shift with each round.
  5. Relationship temperature — Is this a new creator partnership or an established relationship? First-time collaborators need more context and encouragement. Long-term partners can handle more direct notes.
  6. What the creator got right — Did anyone on the internal team call out positives? If the raw feedback is all negative, ask: "What did the creator do well? I need at least one specific positive to lead with."
  7. Non-negotiables vs. nice-to-haves — Which feedback items are mandatory changes (brand safety, legal compliance, factual errors) and which are preferences (pacing, music choice, caption wording)? If the user does not distinguish, ask.
  8. Preferred channel — Will this be sent via email, Instagram DM, TikTok DM, or a project management tool? Channel determines length and formality.

Fallback questions (ask only what the input and context file do not already cover):

  • "What type of content is this feedback for — a Reel, TikTok, Story set, or something else?"
  • "Is this the first round of revisions, or have they already revised once?"
  • "Which of these notes are must-fix items and which are suggestions?"

Core Principles

  1. Lead with What Works, Then Fix What Doesn't — Every revision request starts by naming specific things the creator did well. Not a generic "great job so far" — a concrete observation like "the product demo in the first 5 seconds is exactly the angle we wanted." Creators who feel their work is respected engage more deeply with revision notes. Creators who feel criticized disengage. Test: if the positive feedback could apply to any creator on any campaign, it is too generic. Rewrite it.
  2. Separate Must-Fix from Nice-to-Have — Internal feedback threads collapse mandatory changes and personal preferences into one undifferentiated list. The creator cannot tell which notes are blocking approval and which are optional suggestions. Every revision request must split feedback into two explicit categories: changes required for approval and suggestions the creator can choose to incorporate. A creator who addresses all must-fix items and skips suggestions should receive approval. If they can't, the item was miscategorized.
  3. Translate, Don't Relay — Internal feedback uses shorthand, acronyms, legal language, and blunt phrasing that was never meant for the creator. "The CTA is buried and the disclosure is non-compliant" becomes "Could you move the call-to-action closer to the hook — ideally in the first 3 seconds — and add #ad to the caption? That keeps us both covered on FTC guidelines." The revision request is a translation layer between your internal team and the creator. Relay the intent, not the exact words.
  4. Be Specific Enough to Act On — "Make it more on-brand" is not actionable feedback. "Use the product name 'Glow Serum' instead of 'your serum' and show the packaging label in at least one frame" is actionable. Every revision note must answer: what exactly needs to change, where in the content it applies, and what the result should look like. If you cannot make a note specific, it belongs in a conversation, not a revision request.
  5. Protect the Relationship Across Rounds — Each revision round carries cumulative emotional weight. Round 1 revisions feel collaborative. Round 2 feels tedious. Round 3 feels adversarial. Adjust tone accordingly: Round 1 is enthusiastic and detailed. Round 2 acknowledges the creator's effort on the first revision and keeps the remaining notes tight. Round 3 — if it happens — must be surgical: only true blockers, maximum appreciation for the creator's patience, and a clear signal that approval is close.

Feedback Translation Framework

Step 1: Categorize the Raw Feedback

Read all internal notes and sort every distinct piece of feedback into one of these categories:

CategoryDefinitionExamples
Brand complianceViolates brand guidelines, uses wrong product names, incorrect claims, competitor products visibleWrong product name, unapproved claims, competitor logo in frame
Legal / FTCMissing or incorrect disclosure, unsubstantiated health claims, rights issuesNo #ad, medical claim without disclaimer, uses copyrighted music
Creative directionDoes not match the brief's creative requirements — wrong format, missing key message, wrong product focusProduct shown too late, key message missing, wrong aspect ratio
Technical qualityAudio, video, or image quality issues that affect usabilityBad lighting, audio too quiet, blurry product shot, wrong resolution
Tone and styleContent tone does not match brand voice or campaign moodToo salesy, too casual for the brand, humor that clashes with brand positioning
SuggestionsOptional improvements that would strengthen the content but are not required for approval"Would be cool if..." pacing preference, music swap idea, caption tweak

Step 2: Deduplicate and Consolidate

Internal threads often contain the same note from multiple people in different words. Merge duplicates:

  • If two reviewers flag the same issue, combine into one note citing the strongest version
  • If reviewers contradict each other (one says "too long," another says "add more detail"), flag the conflict and recommend the user resolve it before sending to the creator — do not pass contradictions to the creator
  • If a note is vague even after reading surrounding context ("this doesn't feel right"), ask the user to clarify before including it

Step 3: Prioritize by Approval Impact

Assign each feedback item a priority:

PriorityLabelRule
P1 — Must FixRequired for approvalBrand compliance, legal/FTC, and any creative direction item the brand has flagged as non-negotiable
P2 — Strongly RecommendedExpected but not blockingTechnical quality issues, tone mismatches that affect brand perception
P3 — SuggestionOptional, creator's choiceStyle preferences, nice-to-have improvements, alternative approaches

Step 4: Translate Each Note

For each feedback item, convert the internal language into creator-friendly language:

Internal note: "CTA is buried. Needs to be in the first 3 seconds. Also, the bottle isn't visible enough." Creator-facing version: "Could you move the product mention closer to the start of the video — ideally within the first 3 seconds? That hook is when most viewers decide to keep watching. Also, if there is a natural moment to show the bottle label up close, that would really help viewers connect the video to the product."

Follow these translation rules:

  • Replace commands with requests ("Move the CTA" becomes "Could you move the call-to-action")
  • Add the reason behind each change — creators comply faster when they understand the why
  • Use time references for video ("at the 0:03 mark") instead of vague placement ("near the beginning")
  • Keep legal/FTC notes factual and helpful, not threatening ("Adding #ad to the caption keeps us both covered on FTC guidelines")
  • Preserve any positive framing the internal reviewer used — amplify it in the creator message

Step 5: Draft the Revision Request Message

Structure the message in this order:

1. Opening — Positive and specific Reference something specific the creator did well. Tie it to the campaign goal or their creative strength.

2. Context — Where the content stands One sentence on the review stage: "We've reviewed the draft internally and have a few notes before we can give final approval."

3. Required changes — P1 items Numbered list. Each item states what needs to change, where in the content, and why.

4. Recommended changes — P2 items Numbered list, clearly labeled as recommendations. Frame as "this would strengthen the content" not "you need to fix this."

5. Optional suggestions — P3 items (include only if 3 or fewer) Brief, low-pressure. Frame as ideas, not requests. If more than 3 suggestions exist, pick the top 3 and drop the rest — too many optional notes overwhelm.

6. Closing — Timeline and support State the revision deadline if one exists. Offer to hop on a quick call or answer questions. End with genuine encouragement.

What NOT to Do

  • Do not include contradictory feedback. If two reviewers disagree, resolve it internally before sending. The creator should never receive "Reviewer A wants more energy, Reviewer B wants it calmer." That is your problem to solve, not theirs.
  • Do not use internal jargon. "Above-the-fold CTA placement" and "non-compliant disclosure" are your language, not the creator's. Translate everything into plain language a creator would naturally use.
  • Do not send more than 7 revision notes in a single message. If the internal review generated 15 separate notes, consolidate by theme. More than 7 distinct items feels like a teardown, regardless of how nicely they are phrased.
  • Do not skip the positive opening. Even if the content needs significant rework, there is always something the creator did right. Find it. Lead with it. Skipping the positive opener tells the creator you only see what is wrong.
  • Do not turn suggestions into demands. If the team's note was "it would be cool if she used the trending audio," that is a suggestion. Do not convert it to "Please replace the audio with [trending sound]." Preserve the optionality.
  • Do not include passive-aggressive language. "As we mentioned in the brief" and "per our previous conversation" sound accusatory even when intended as neutral. Just restate the requirement cleanly.

Tone Calibration by Round and Channel

By revision round:

RoundToneStructureLength
Round 1Warm, collaborative, detailedFull structure: positives + all notes + suggestions250-400 words
Round 2Appreciative, focused, conciseBrief positive noting improvement + remaining items only150-250 words
Round 3Surgical, grateful, almost doneOnly P1 blockers + clear signal that approval is imminent75-150 words

By channel:

ChannelAdjustments
EmailFull structure, headers allowed, professional but warm
Instagram DMShorter, no headers, conversational, break into multiple messages if needed
TikTok DMBrief, casual, direct — TikTok DMs are not the place for 400-word revision notes
Slack / project toolStructured, can use formatting, balance between email and DM

Segment-Aware Guidance

Adjust the revision request based on who is using it:

SMB brands (solo marketer, under 50 creators)

  • The user is probably the only reviewer, so "internal notes" may just be their own thoughts jotted down — everything is manual and they are juggling 15 other tasks. Help them structure those thoughts into a professional message.
  • Keep the output short and sendable — this person does not have time to edit a 500-word revision request.
  • Emphasize the relationship angle — SMB brands cannot afford to burn a creator relationship over a picky revision.

Mid-Market brands (influencer team, 50-200 creators)

  • Multiple reviewers are common. Deduplication and conflict resolution become critical.
  • The person sending the feedback may not have written it — they are translating their team's notes. Help them own the message as if the feedback came from one unified voice.
  • Include a brief internal summary alongside the creator message so the sender can track what was communicated.

Enterprise brands and agencies (dedicated team, 200+ creators)

  • Feedback may come from brand managers, legal, creative directors, and agency account leads — each with different priorities. Consolidation is the primary value.
  • Format for handoff: the person running the review tool may not be the person sending the message. Make the output clear enough that anyone on the team can send it.
  • Agencies: confirm which client's voice the message should reflect. Agency voice and client brand voice are not the same.

Output Format

Structure every output with these sections:

1. Feedback Summary (Internal Reference)

A quick-scan table for the user's records showing what was captured from the raw notes:

| # | Feedback Item | Category | Priority | Source |
|---|---|---|---|---|
| 1 | Product name incorrect — called it "the serum" instead of "Glow Serum" | Brand compliance | P1 — Must Fix | Marketing lead |
| 2 | No FTC disclosure in caption | Legal / FTC | P1 — Must Fix | Legal review |
| 3 | Product shown at 0:12 — should be in first 3 seconds | Creative direction | P1 — Must Fix | Creative director |
| 4 | Background music too loud, hard to hear voiceover | Technical quality | P2 — Recommended | Brand manager |
| 5 | Could try the trending "get ready with me" format | Suggestion | P3 — Optional | Social team |

2. Conflict Flags (If Any)

If reviewers contradicted each other, list the conflicts with a recommended resolution:

CONFLICT: Marketing says "keep it under 30 seconds" but Creative says "add more product detail."
Recommended resolution: Prioritize the 30-second limit — suggest the creator show the product in a quick unboxing montage rather than extended demo.

3. Revision Request Message (Ready to Send)

The complete, copy-paste-ready message formatted for the specified channel. Follows the structure from Step 5 above.

4. Revision Checklist (When Helpful)

For complex revisions with 4+ items, include a simple checklist the creator can reference:

Revision checklist:
[ ] Use product name "Glow Serum" (not "the serum") — at least twice in voiceover
[ ] Add #ad to the caption
[ ] Show product in first 3 seconds
[ ] Lower background music volume so voiceover is clear

Approximate output length: 400-800 words depending on feedback volume and revision round.

Quality Check

Before delivering the formatted feedback, verify:

  1. Every P1 item from the raw notes appears in the revision request — No required change was dropped during translation. Cross-check the internal summary table against the creator message.
  2. No contradictory feedback made it into the message — If two conflicting notes exist, they are flagged in the Conflict section, not passed to the creator. The revision request speaks with one voice.
  3. Each revision note is specific enough that the creator can act on it without asking a follow-up question — "Make it more on-brand" fails this test. "Use the product name Glow Serum instead of 'your serum' and show the packaging label" passes.
  4. The message starts with a genuine, specific positive — Read the opening line. Does it name something the creator actually did well in this specific draft? If it could apply to any creator on any campaign, rewrite it.
  5. A creator partnerships manager reviewing 8 content drafts in one afternoon would send this message to the creator within 5 minutes of reading it — If the output needs significant rewriting before it is sendable, it is not useful enough.

Related Skills

  • If you need to build the original content brief before the creator starts production, see creator-content-concept-generator
  • If you need to check submitted content against every brief requirement with a structured pass/fail, see content-to-brief-compliance-checker
  • If you need to review a caption or script specifically for FTC disclosure compliance, see ftc-disclosure-spot-checker
  • If you need to write the outreach message that starts the partnership, see creator-outreach-sequence-generator
  • If you need to summarize agreed terms from a negotiation thread, see verbal-agreement-summarizer
  • If you need to follow up on late or missing content deliverables, see universal-creator-follow-up-chaser

平台分布

Codex

34.69%
按下载量换算111

Claude

28.92%
按下载量换算93

Cursor

19.56%
按下载量换算63

Gemini CLI

10.2%
按下载量换算33

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills