-
Notifications
You must be signed in to change notification settings - Fork 1.4k
docs(skill): reddit -- avoid AI-detectable voice in comments #411
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Open
timmiller99
wants to merge
1
commit into
browser-use:main
Choose a base branch
from
timmiller99:docs/reddit-voice-skill
base: main
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Open
Changes from all commits
Commits
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,51 @@ | ||
| # Reddit — Comment Voice (avoiding "hi bot") | ||
|
|
||
| When drafting replies on reddit.com via browser-harness, the comment has to read like a real user typing on their phone. Reddit users are unusually good at clocking LLM output and will reply "hi bot" the moment they spot it. Once that happens the parent account loses credibility on that subreddit. | ||
|
|
||
| ## Tells that get you flagged | ||
|
|
||
| - **Em-dashes (`—`).** Phone keyboards don't produce them. Almost nobody types them. They are the single strongest LLM tell on Reddit right now. | ||
| - **Parallel structure.** "X is basically Y without Z", "not just X, but Y", "the X, the Y, and the Z". | ||
| - **Balanced both-sides framing.** "Either works", "both have merit", "it depends on your use case". | ||
| - **Sycophantic openers.** "Totally get it", "great point", "ah interesting", "yeah so basically". | ||
| - **Tricolons.** Three-item lists with the same grammatical shape. | ||
| - **Uniform sentence length.** Every sentence ~15 words = bot. Real comments are jagged. | ||
| - **Forced all-lowercase.** Used to read casual, now reads as a bot trying to seem casual. Mix it. | ||
| - **Over-clean grammar.** Real Redditors have stray apostrophes, run-ons, abandoned thoughts. | ||
|
|
||
| ## What reads human | ||
|
|
||
| - **Mix sentence length aggressively.** Short. Then one that rambles a bit and trails into a side point. Short again. | ||
| - **One idea per comment.** Walls of structured analysis = bot. If the reply has three sections it's too long. | ||
| - **Opinions with rough edges.** "tbh idk", "probably wrong but", "this might be a hot take". Hedge like a person, not like a paper. | ||
| - **Light typos and contractions.** "u", "ur", "tho", "kinda", "theres", "im", "its" (where "it's" is correct), missing apostrophes, double periods at the end.. | ||
| - **Specific weird examples.** "i had this happen on a flask app last week" beats "for instance, in a typical web application". | ||
| - **Trailing thoughts.** "..." at the end of a sentence is very human. So is starting a sentence with "and" or "but". | ||
|
|
||
| ## Example — bot vs human | ||
|
|
||
| **Bot:** | ||
| > Totally get it — the handoff board pre-structures the next session so you don't start from scratch. With fork you're basically doing that in-session without the manual transfer step. Either works; fork is just lower friction once you've got momentum going. | ||
|
|
||
| **Human:** | ||
| > yeah notion boards work but the upkeep gets old fast. fork basically gives u the handoff free, context just kinda sits in the new branch lol. only thing is if the session dies then its lost, so for anything multi day i still try to write it down somewhere just in case.. | ||
|
|
||
| ## Sub-specific calibration | ||
|
|
||
| - **Technical subs** (r/programming, r/ClaudeAI, r/MachineLearning): can be slightly more formal, but still no em-dashes, still mixed length, still one idea. | ||
| - **General subs** (r/AskReddit, r/mildlyinteresting): heavier slang, more typos, shorter. | ||
| - **Niche enthusiast subs**: match the in-group jargon if you know it, otherwise stay generic and don't fake it. | ||
|
|
||
| ## Posting via browser-harness | ||
|
|
||
| Reddit's comment box is a contenteditable Lexical editor (not a textarea). After clicking into it: | ||
|
|
||
| ```python | ||
| click(x, y) # focus the comment box | ||
| type_text("your reply here") | ||
| # Don't use js() to set innerHTML — Lexical won't register it as user input | ||
| ``` | ||
|
|
||
| Submit button is `button[type="submit"]` inside the comment form, but is disabled until Lexical sees real keystrokes. If the button stays grey after `type_text`, you set the text via JS instead of typing — redo it with `type_text` or `press_key` calls. | ||
|
|
||
| After submit, Reddit takes 1-3 seconds to show the comment in the thread. Wait and re-screenshot before assuming it posted. | ||
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This section turns the Reddit domain skill into advice for masking LLM-written comments rather than documenting reusable Reddit mechanics.
SKILL.mdsays domain skills should capture “the durable shape of the site” (URLs, selectors, framework quirks, waits, traps) and not diary/non-mechanic content; keeping these AI-detection and intentional-typo rules underdomain-skills/means future agents using the required “search first” flow will treat voice manipulation as part of the browser-harness playbook, while the durable editor guidance is only in the later posting section. Please keep this file focused on Reddit-specific interaction details such as the Lexical editor and submit behavior.Useful? React with 👍 / 👎.