agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/.claude/skills/log-twitter-niche-map-region/SKILL.md

log-twitter-niche-map-region

Skill63 starsChanged 3 days agoMITView on GitHub

A new region of the niche map, then outbound to it. A log play (docs/useCases/twitter/nicheMapRegion.md): The niche map is the owner's network ([networkMaxing.md](networkMaxing.md)). A region is one world added to it, for example the YC 2026 batches (the founders, their companies and the partners around them). The region's people go through the same niche map pipeline as everyone else: three months of who talks to them, everyone scored, bands set, and each person's closeness to the owner (ring 1 engaged with him, ring 2 talks to his followers, ring 3 follows him, ring 4 new). The outbound then goes to ring 2 first, the warmest people who haven't talked to him yet.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    The owner names the batches, the areas and the money: Never add a world, a batch or a source he didn't say (rule 39).

    line 43

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

A new region of the niche map, then outbound to it

This is the log play twitter/nicheMapRegion. Before anything else, call the log MCP tool getGuide({ useCase: "twitter/nicheMapRegion" }): it returns this page and makes it the session's play, so every log tool's answer says which step you are on and a send with no play loaded is refused (rules 157, 158). Then follow the steps in order.

Draft every message from that person's own thread and matt's real replies (voice with no text, search), never from memory, and never send two that read alike (sameAsSent).

The niche map is the owner's network (networkMaxing.md). A region is one world added to it, for example the YC 2026 batches (the founders, their companies and the partners around them). The region's people go through the same niche map pipeline as everyone else: three months of who talks to them, everyone scored, bands set, and each person's closeness to the owner (ring 1 engaged with him, ring 2 talks to his followers, ring 3 follows him, ring 4 new). The outbound then goes to ring 2 first, the warmest people who haven't talked to him yet.

How the owner asks for it: "for getting accepted into YC ... make a new area of the niche map through the systems we've already made ... the YC world and the funds' world, based on the 26 batch", "the YC stuff in and around the boat.dev people and the people involved in GTM tools", "pull all of my new followers ... set a budget of like 60$", "start with the 43", "not just treg but the actual index that we did already".

Run it from a fresh context

Say: "go on docs/useCases/twitter/nicheMapRegion.md for <the world>, <the areas>". The agent reads this page and README.md, loads getGuide({ topic: "xSending" }), and runs the steps.

The steps

#StepToolDone when
1The region's people from a free public source. For YC: every company page of the batches the owner named (founders, their X handles, the company account, the group partner), pulled slowly. YC's team page gives the partnersthe niche map repo's 78_yc_batch.py (REPOS/whop/research/nicheMap), YC's public pagesdata/yc/companies.json, people.json, partners.json
2Keep only the owner's area: the companies whose one-liner, description or tags fit what he named (GTM tools, the agent-infrastructure world around a company like Boat). Hardware, biotech and the rest are outa filter over the batch files (data/yc/focus.json), checked by handthe region's people (focusPeople.json) and the circles to pull (circles.json)
3Mark who already engages with the owner (his notifications store) and who he already talks to (the inbox), so a warm person never gets a cold templatethe account's notifications.json and inboxSync.jsoneach person's closeness and engagement
4Say the price before the paid run, and run only to the budget the owner set (rule 41). His new followers can go in the same runthe ledger (data/run_ledger.json), GG_BUDGETthe owner said yes to a number
5The region through the niche map pipeline, with the region's circles beside his followers: their three months of mentions, the new people's profiles, scores, own posts, closeness and bandsrun_new_followers.sh with GG_CIRCLES_FILE=data/yc/circles.json. API-heavy pulls go to Boat sandboxes with the owner's treg login when his machine is under load; the rest reads the cache locallydata/final_scores.json holds the region, with rings and bands
6The outbound list: the region's go getters (gates passed; band member, collaborator, sponsor or affiliate) in ring 2 first, then ring 2 around themdata/yc/goGetters.jsonthe list, ranked
7One recent post of each person's own to open on: the index holds their posts, and one fresh search each confirms they posted this week. Nothing new: they waitthe index (search_index.json, search_posts.jsonl), treg's posts searcheach with a user id and a post from the last few days
8Whether they follow the owner and he follows themtwitterFollowsthe relationship, for the opener's wording
9Each opener the cold DM way (engagerDms.md): their tweet first, then one line about it, never an askthe agentone draft per person
10Queue them as dms and check them: house style, and nobody he already talks to on X ChattwitterQueueBuild { action: "dm" }, twitterQueueCheckall pending, none dropped or failed
11Gate every opener with the voice gate (his writing across all his conversations), rewriting until it passesvoice { to, context, text }every draft passes
12Follow, like the post, then the opener, a few people a day, minutes apart (rule 94)twitterFollowLive, twitterLikeLive, the queue runX's answer is sent for each

What to know

  • The owner names the batches, the areas and the money. Never add a world, a batch or a source he didn't say (rule 39).
  • The page reads YC slowly, about one page every 1.5 s, and keeps every page it read in data/yc/pages, so a rerun costs nothing.
  • A dropped write is never retried through. A 429 or a soft block stops the sends, and the next action waits for the owner (rule 94).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.