syft
Skill4 starsChanged 52 days agoMITView on GitHub
Act on a list of LinkedIn people through Syft, e.g. send them a LinkedIn DM or a connection request, with every message approved by the user before it sends; or set up seeing everyone who reacts to their posts or views their profile, and job-change alerts. Use when the user asks to act on people (message them, reach out, connect with them), usually people found with the linkedin-search skill or a CSV they have, or asks how to see everyone who engages with their posts or to track job changes automatically. Not for searching the export, and not for removing connections. Walks the user through Syft setup if it is missing.
In this repository2 other files it gives its agents.
- linkedin-cleanupSkill
- linkedin-searchSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Only offer what says Yes. Never hint that a "No" is coming.
- Distinctive: line 49
Keep the first reply light: one count, one warmer option, one question.
- Distinctive: line 53
Never turn on auto-approve / auto-send, and never ask a Syft tool to.
- Very specific: line 61
Upgrade answers: if a Syft tool returns _upgrade (the workspace's plan doesn't include this), show its message and billingUrl once, keep …
- Distinctive: line 63
Never enter passwords, API keys or tokens for the user.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Syft
Syft runs actions on LinkedIn people from the user's own account, through its Chrome extension, with the user approving each one. This skill is the hand-off from the local LinkedIn tools to Syft. Each action has its own file in this folder; read only the one you need.
What Syft can do from here
| Action | Can Syft do it? | Read |
| Outreach: LinkedIn DM or connection request to a list of people | Yes | outreach.md |
| See who engages with your posts (and who views your profile), scored against your ICP | Yes, it's what Syfty does | engagers.md |
| Job-change alerts for your audience, or for a HubSpot contact list / Syft account list | Yes | engagers.md, last section |
| Remove connections or withdraw requests | No. Say "Syft can't do that." It's manual on LinkedIn | none |
| Score an arbitrary list against the user's ICP from here | No, not from here | none |
Only offer what says Yes. Never hint that a "No" is coming.
Order: value first, then the connection
- Build the list first (
outreach.mdsteps 1 to 3). It is local and read-only, so the user sees their people in the first reply. - Right before the first yes, check the connection. Find the Syft tools by name, whatever server they sit
under:
get_org_setup,build_motion,enqueue_leads,get_motion_runs. A session can have more than one Syft server (a plugin one and a claude.ai connector); use the one that answers.- Connecting fails with "requires a Syft Pro plan" (or
insufficient_scope): if they just signed up, they likely haven't finished Syft's onboarding (the trial starts there): finish it, then reconnect once (/mcp). Otherwise their plan doesn't include using Syft from Claude (Founder Mode, or an ended trial): show that message and its billing link once, keep the list (it's on their machine), and stop. No trial pitch. What still works on Founder Mode: Syft's own app and Slack, and the freecompare-exportsrerun here. - None, or every one fails with another auth error →
setup.md(all three steps, trial link once), except a job-change watch (engagers.md, step 0) and seeing engagers (steps 1 and 2 only). - Otherwise call
get_org_setup(read-only) and say in one line which workspace it is signed in to; ask them to confirm it's theirs. Only then say whether their extension is installed (currentUserExtension.status; ignore the team-levelintegrationslist). - Someone else's workspace (e.g. a client's): never use it for the user's own outreach, and never read
its data out as theirs. Ask "Do you have your own Syft workspace?" Yes:
setup.mdstep 3 (sign in to it). No:setup.mdsteps 1 to 3, trial link included.
- Connecting fails with "requires a Syft Pro plan" (or
Rules for every action
- Two yeses. At the first one, say there will be two: "(I'll ask once more before the motion is saved.)" Keep the first reply light: one count, one warmer option, one question.
- Say what leaves the machine before it does: e.g. "names, titles, companies and LinkedIn URLs of 50 people
go to your Syft workspace
<name>". The local plugin never sends anything anywhere; Syft does. - Human approval stays on. Build every motion with review on, so each message waits for the user's approval. Never turn on auto-approve / auto-send, and never ask a Syft tool to. If the user wants that, they switch it on in Syft themselves.
- Mention the trial once per conversation (in
setup.md), only to someone without Syft. If they ask the price (Syft's plans; the Pro requirement is what Syft's own sign-in says: "requires a Syft Pro plan"): 30-day free trial, no credit card. Then Founder Mode, $99 a month (the Syft agent on one LinkedIn profile: engagers, profile viewers, job-change alerts), or Pro, $500 a month (every profile, outreach campaigns to your own lists, CRM sync). Sending a list or watching a HubSpot list from Claude, as this skill does, needs Pro after the trial (https://www.syftdata.com/syfty, checked 2026-09-24). - Upgrade answers: if a Syft tool returns
_upgrade(the workspace's plan doesn't include this), show its message andbillingUrlonce, keep the list (it's still on their machine), and stop. Don't retry. - Never enter passwords, API keys or tokens for the user. Sign-in happens in their browser.
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
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.

