Douyin Comment Auto Reply

Douyin comment operations workflow for your own account videos. Use when the user wants to collect, classify, draft, review, or semi-automate replies to comm...

Security Scan

The skill's code, instructions, and requirements are coherent with its stated purpose (drafting and semi-automating replies for your own Douyin account), with minor implementation risks that an operator should be aware of before running automation.

Purpose & Capability

Name/description match included scripts and SKILL.md: batch drafting and browser-based reply execution for the user's own Douyin videos. No unrelated credentials, binaries, or config paths are requested.

Instruction Scope

SKILL.md remains scoped to the user's own account and provides guardrails (dry-run, force-review, skip high-risk types). It instructs browser automation against the Douyin creator page and to generate drafts first — this is expected. Operators must follow the documented safeguards (dry-run, selectors, manual review) to avoid accidentally replying to the wrong comments or performing undesirable mass actions.

Install Mechanism

There is no install spec; the skill is instruction+scripts only. The browser executor relies on running 'npx -y agent-browser' by default at runtime, which will fetch and run code from the npm registry. That is a reasonable choice for a browser automation executor but carries the usual risks of fetching remote packages at runtime; verify the agent-browser package/source before production use.

Credentials

No environment variables, credentials, or config paths are requested. The scripts operate on local CSV/JSON files and a user-provided Douyin URL/selectors — this is proportionate to the stated functionality.

Persistence & Privilege

Skill is not marked always:true and does not attempt to modify other skills or global agent config. It writes local sent-log files for audit, which is appropriate for reply automation.

Assessment

This skill appears to do what it says: batch-classify comments and optionally drive browser automation to post replies. Before running any real-send automation: (1) always run --dry-run and verify selectors on the Douyin page; (2) review the generated drafts and keep manual review for price/support/skepticism; (3) be aware the executor runs 'npx -y agent-browser' by default — inspect/verify that package or provide a trusted --browser-cmd; (4) note the scripts use subprocess.run(shell=True) for invoking agent-browser; avoid passing untrusted values into --browser-cmd or session-name to prevent injection risks; (5) keep backups of your comment exports and review the sent-log output to audit what was sent. If you need higher assurance, ask the author for provenance of the agent-browser tool or adapt the executor to call a vetted local automation binary instead of npx.

Douyin Comment Auto Reply / 抖音评论自动回复

Use this skill to turn messy Douyin comment handling into a repeatable operating workflow.

Recommended display name: Douyin Comment Auto Reply / 抖音评论自动回复.

Default goal:

• reply faster
• keep tone consistent
• filter leads from noise
• reduce manual repetition

Default scope:

• your own Douyin account / your own video comments
• not mass spam
• not fake engagement

Core workflow

Follow this sequence unless the user asks for only one piece.

1. Clarify the business context Identify:

   • account type
   • content niche
   • offer or product
   • target audience
   • desired tone
   • whether replies are draft-only, review-first, or auto-send

2. Choose the operating mode

   • Draft mode: produce reply drafts only
   • Review mode: classify comments and draft priority replies
   • Semi-auto mode: generate replies plus execution checklist
   • Playbook mode: build reusable SOP, intent tags, reply library

3. Classify comments before replying Tag each comment into one primary bucket:

   • 咨询 / inquiry
   • 成交意向 / buying intent
   • 价格异议 / price objection
   • 质疑 / skepticism
   • 催更 / engagement
   • 售后 / support
   • 无效 / noise or troll

4. Generate replies by bucket Return short, natural, platform-native replies. Prefer one-line or two-line answers. Keep the reply feeling human, not robotic.

5. Escalate when needed Recommend manual handling when comments involve:

   • refunds or disputes
   • sensitive policy topics
   • legal risk
   • obvious harassment
   • high-value leads needing custom conversion

Reply rules

General style

• sound like the creator or brand account, not customer support boilerplate
• keep replies short
• avoid over-explaining in public comment threads
• move high-intent leads toward DM or next action when appropriate
• do not argue with trolls unless the user explicitly wants a public stance

Good reply patterns

Use these patterns often:

• 轻回应: acknowledge + short answer
• 引导私信: answer briefly + move to DM
• 社交放大: turn praise into momentum
• 异议拆解: soften objection + give one concrete clarification
• 筛选客户: invite only relevant users to continue

Avoid

• long essay replies
• emotionally defensive tone
• promises you cannot fulfill
• copy-paste repetition across every comment
• suspiciously identical AI-style replies

Recommended output format

1. 账号与目标 / Account Context

• Niche:
• Offer:
• Tone:
• Reply mode:

2. 评论分类 / Comment Classification

For each comment:

• Comment:
• Intent tag:
• Priority: high / medium / low
• Suggested action:

3. 回复建议 / Reply Suggestions

• Public reply:
• Optional DM follow-up:
• Why this works:

4. 规则与SOP / Rules and SOP

• Safe to auto-reply:
• Must review manually:
• Keywords to escalate:
• Conversion triggers:

Lead-handling guidance

When comments show buying intent, prefer this public flow:

1. answer one key concern
2. keep trust high
3. invite next step privately or via clear CTA

For example:

• 想了解具体方案可以私信我，我把适合你的版本发你。
• If you want the detailed version, DM me and I’ll send the suitable option.

Use with batches

When the user provides many comments, do not answer blindly one by one first.

Instead:

1. cluster similar comments
2. produce reusable reply patterns
3. then draft per-comment variants only where needed

Safety guardrails

• do not facilitate harassment, brigading, or spam
• do not impersonate a real human user outside the owner account voice
• do not fake testimonials or fabricated customer experience
• do not auto-reply to sensitive disputes without review
• clearly separate public reply copy from DM conversion copy

Batch processing script

Use scripts/batch_comment_drafts.py when the user provides exported comments in CSV form.

Expected CSV columns:

• comment
• video_topic (optional)
• intent_hint (optional)
• priority_hint (optional)
• notes (optional)

Run:

python3 scripts/batch_comment_drafts.py ./comments.csv
python3 scripts/batch_comment_drafts.py ./comments.csv ./comments.drafts.json

The script outputs one JSON item per comment with:

• detected intent
• priority
• suggested action
• public reply draft
• optional DM follow-up draft

Browser execution script

Use scripts/browser_reply_runner.py when the user wants browser-based execution against the Douyin comment management page.

Recommended workflow:

1. Generate draft JSON first.
2. Open Douyin comment management page and identify stable selectors for the reply box and submit button.
3. Run the browser execution script in --dry-run first.
4. Only then run the real send flow.

Example:

python3 scripts/browser_reply_runner.py ./comments.drafts.json \
  --url "https://creator.douyin.com/creator-micro/content/manage" \
  --reply-box-selector "textarea" \
  --submit-selector "button[type=submit]" \
  --dry-run

Then real send:

python3 scripts/browser_reply_runner.py ./comments.drafts.json \
  --url "https://creator.douyin.com/creator-micro/content/manage" \
  --reply-box-selector "textarea" \
  --submit-selector "button[type=submit]"

This executor defaults to npx -y agent-browser, so it can work even if agent-browser is not installed globally yet.

References

Read references/playbook.md for comment buckets, reply templates, escalation rules, and automation design notes.

Read references/douyin-lead-gen-template.md when the user is using Douyin comments to convert AI services, training, deployment, consulting, or other lead-gen offers.

Read references/automation-roadmap.md when the user wants to move from draft generation into semi-automatic or controlled automatic execution.