O que dev-comment faz?
Turns a reading into a dev.to comment that adds concrete value. dev.to comments are threaded (max depth then flatten), support Markdown including code blocks, and support only the ❤️ reaction on comments. Comment count + comment likes feed into the post's "hotness" score.
Before you write
- Read the article. Identify one specific thing to add: a correction, a working snippet, a citation, a focused follow-up question, or a relevant personal experience.
- Check the user's voice in
~/.claude/devto/voice.md. - dev.to Code of Conduct applies: welcoming, respectful of differing views, graceful acceptance of criticism. No hate speech, trolling, insults, harassment, doxxing, or plagiarism.
The shape
Line 1 the point. Lead with it. One sentence.
Body short, 1-3 lines per paragraph. Code blocks for snippets.
Every snippet needs context text before it.
Close one focused question or one thank-you. Never both.
Tone warm, concise, evidence-based. Lowercase-casual is fine.
Length: under 200 words is the working range. dev.to comment threads flatten after max depth — to continue, reply to the last comment in the overall thread.
The loop
1. Pick three angles, not one. Different approaches to the same article:
- correction/addition ("Actually, X also handles Y…")
- working snippet ("Here's a minimal reproduction…")
- follow-up question ("Did you consider Z for this case?")
- personal experience ("I hit this same issue with…")
2. Draft the full comment on the strongest angle.
3. Humanize. Run through /dev-human before showing it.
4. Print the block. Copy-ready:
COMMENT READY
angle: #3 follow-up
length: 85 words
humanizer: 3 artefacts stripped, human score 88 PASS
[comment body]
Reply "yes" to log it, or tell me what to change.
5. Never post automatically. dev.to's public API (v1) exposes comment
read-only (GET /api/comments?a_id=, GET /api/comments/{id}). There is no
comment-create endpoint. The user pastes it into the dev.to comment box
manually. On "yes", append to ~/.claude/devto/log.md.
Rules that make the difference
- Add concrete value. Cite sources, share a working snippet, correct with specifics, ask a focused follow-up. No "Great post!" drive-bys.
- Lead with the point. Don't bury the lede.
- Code blocks for snippets. Every snippet needs intro context.
- Match dev.to's Code of Conduct tone. Welcoming, respectful, graceful with criticism.
- Disclose AI assistance where applicable (per dev.to AI guidelines).
- No AI-generated comments. dev.to policy: no bots / AI-generated comments (exceptions: basic translation/grammar tools, assistive tech).
- Subscribe to the post's comments if you want to catch replies.
- Hide, don't argue. If someone is abusive on your post, hide it via the dropdown rather than escalating.
- Reply to your own post's comments. Authors who respond build discussion. Prompt replies (especially early comments) boost thread activity.
- #discuss tag on your own posts for open-ended questions — can earn "Top Discussion of the Week" badge.
Example
/dev-comment this post about React server components
ANGLES
1. Correction "Server Components don't replace Client Components; they complement them."
2. Snippet "Here's a minimal 'use client' boundary example…"
3. Question "How do you handle data fetching at the boundary?"
Shipping #2: gives the reader something they can paste and run.
[comment body with code block]
COMMENT READY
angle: #2 Snippet
length: 92 words
humanizer: 2 artefacts stripped, human score 85 PASS
Self-improvement: angle learning
After publishing comments, note which angle type (correction / snippet /
question / experience) draws the most replies on the user's own posts.
Promote that angle in the 3-angle options; demote angles that consistently
get zero replies. Track in ~/.claude/devto/log.md if you want per-comment
analytics (comment replies are not exposed by the Forem API, so this is
manual observation).
Skipped (ponytail)
- No comment-create automation (Forem API limitation).
- No reaction toggling via API (disabled in some MCP servers due to 401 scope; web UI only).