lnvestor/twitr-skills / skills/x-presence
x-presence
VerifiedRun an unattended X/Twitter presence — watch topics and mentions in real time, draft posts and replies in the user's own voice, publish original posts on a schedule, and hold replies for a one-tap approval. Trigger whenever the user wants to automate their X/Twitter account, grow an audience on X, schedule or draft tweets, set up an X routine or cron job, get alerted about topics or mentions on X, or run an X agent in the background — including casual phrasings like "post for me on X" or "keep an eye on Twitter for me". Works in Claude Code routines, Hermes cron, and OpenClaw cron. Paid per call in USDC via twitr.sh through an x402 wallet — no API key, no signup. Do NOT trigger for reading a single tweet or profile (use x-twitter-data), for bulk dataset exports (use x-twitter-research), or for non-X platforms.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.2 KB |
| references/ranking-signals.md | markdown | 5.7 KB |
| references/scheduling.md | markdown | 4.9 KB |
| references/x-rules.md | markdown | 3.1 KB |
references/ranking-signals.md
View on GitHub# What actually ranks on X
Read this when the user asks what grows an account, or pushes for follow/like automation.
## The published weights
X open-sourced its ranking stack. The heavy ranker predicts ten engagement probabilities and
combines them as a weighted sum. These weights are **verbatim** from
[the-algorithm-ml/projects/home/recap/README.md](https://github.com/twitter/the-algorithm-ml/blob/main/projects/home/recap/README.md),
dated **April 5 2023** in the document itself:
| Signal | Weight |
|---|---|
| `reply_engaged_by_author` — you reply, the author replies back | **75.0** |
| `reply` | 13.5 |
| `good_profile_click` — opens your profile *and* likes/replies | 12.0 |
| `good_click` — opens the conversation and engages | 11.0 |
| `good_click_v2` — opens the conversation, stays 2+ minutes | 10.0 |
| `retweet` | 1.0 |
| `fav` (like) | 0.5 |
| `video_playback50` | 0.005 |
| `negative_feedback_v2` — "show less often", block, or mute | **−74.0** |
| `report` | **−369.0** |
## Read these honestly
**Treat them as direction, not arithmetic.** Three caveats matter, and the skill should not
overstate this:
1. **They are dated April 2023.** The same README says the weights are "not hardcoded in the
code" and can be adjusted at any time. The current repo confirms it — the serving params hold
no numeric constants.
2. **X deliberately jitters them.** `HomeGlobalParams.scala` exposes `AddNoiseInWeightsPerLabel`
and `EnableDailyFrozenNoisyWeights`. Micro-optimising against exact values is futile by design.
3. **The famous multipliers are not Twitter's.** "A reply is worth 150 likes" was derived by a
blogger dividing 75.0 by 0.5 the day the repo dropped, and it has been recycled ever since —
including by sites presenting the 2023 numbers as current "Phoenix" ranker constants. The
real current weights live in a params module X did not publish. **Do not quote multipliers as
if X stated them.**
What survives all three caveats is the **ordering**, and that's enough to act on.
## What follows from the ordering
- **Conversation beats consumption.** Every high-weight signal involves someone replying,
clicking through, or dwelling. Likes and retweets sit near the bottom.
- **The best single outcome is a reply the author answers.** So reply to people who actually
converse — not megaphone accounts that never respond.
- **Negative feedback is catastrophic and asymmetric.** One report outweighs hundreds of likes.
This is the mechanical case for conservative behaviour: a single spammy blast can undo weeks.
- **Follows and likes are not in the objective function.** Automating them buys nothing the
ranker rewards, while risking the signal that destroys reach. The prohibited tactics are also
the ineffective ones.
- **Profile matters.** `good_profile_click` at 12.0 means someone visiting your profile and then
engaging is heavily rewarded — so bio and pinned post do real work.
## Pipeline facts worth knowing
From [the-algorithm/home-mixer/README.md](https://github.com/twitter/the-algorithm/blob/main/home-mixer/README.md)
and the repo root README:
- Candidate generation → ~6000 features hydrated → ML scoring → filters → mixing.
- **~50% of For You posts come from the in-network search index** (people who follow you). So
followers still gate half your distribution — real follower growth compounds, which is why
it's worth earning rather than automating.
- **RealGraph** predicts how likely user A is to interact with user B, and a graph-feature
service tracks things like how many of A's following liked B's posts. Consistent engagement
with one community compounds; scattershot engagement doesn't.
- Explicit filters include **author diversity** (you can't dominate one feed) and **feedback
fatigue** (the same author repeatedly gets downranked). Posting more is not linearly more
reach.
## Three systems, don't confuse them
Most bad X advice comes from collapsing these. Keep them separate and the guidance stays
defensible:
1. **The ranking code** — what scores a post. Weights, filters, candidate sources.
2. **The written rules** — what gets an account actioned. A rule violation is not a score.
3. **The monetization program** — what gets a post demonetized. Demonetization is not downranking.
Two concrete examples of the confusion. **Deleting posts** carries no ranking penalty — there's no
account-level engagement-average feature in the code, and X *requires* deletion to lift
enforcement — but "posting and deleting the same content repeatedly" is a written-rule violation.
**Duplicate content** ("copypasta") is likewise named in the rules, yet there is no text-similarity
ranking filter; dedup in the code is by tweet id and conversation id only. Both matter — as
policy, not as score.
## Official rate limits
Verified from X's own limits page: **50 original posts and 200 replies per day** for unverified
accounts. Treat these as the ceiling that must never be approached, not a target. (The old
"2,400 updates/day" figure still on that page is stale.)
## Premium, accurately
X documents "a slight preference for replies from verified accounts" and tiered reply
prioritization, and says it is "currently testing the levels." That is **replies and
conversations only**, of unstated magnitude. The widely-quoted 2–4× *For You* multiplier for
Premium does not exist in the code. Don't promise a reach boost from a subscription.
## The strategy this implies
1. Reply substantively in a narrow topic, to people who reply back.
2. Post original content on a steady cadence — permitted, and it feeds the in-network half.
3. Make the profile convert a visit into a follow.
4. Never earn a mute, block, or report.
Which is exactly what this skill automates, and exactly what it refuses to do.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |