Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add matteolegrottaglie/curtis --skill linkedin-outreachgit clone --depth 1 https://github.com/matteolegrottaglie/curtisWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/matteolegrottaglie/curtis/linkedin-outreach)<a href="https://agentmods.dev/skills/matteolegrottaglie/curtis/linkedin-outreach"><img src="https://agentmods.dev/badge/skills/matteolegrottaglie/curtis/linkedin-outreach/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/matteolegrottaglie/curtis/linkedin-outreach"><img src="https://agentmods.dev/badge/skills/matteolegrottaglie/curtis/linkedin-outreach.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00055 | $0.00962 |
| Opus 5 | $0.00028 | $0.00481 |
| Sonnet 5 | $0.00011 | $0.00192 |
| Haiku 4.5 | $0.00006 | $0.00096 |
Grade A, and why
linkedin-outreach scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
LinkedIn outreach with Curtis
The MCP server already carries its own manual in its instructions (operation order, limits,
handling a HALT). This skill covers what those instructions cannot say: how to prepare a campaign
that actually works, and how to read the result.
Preparing the list
Before importing, look at the CSV. Only the profile-URL column is required, but the quality of the others decides the quality of the messages:
- A
first_nameseparate fromlast_namemakes{firstName}reliable. With onlyfull_namethe tool takes the first word: on "Dr. Mark White" that becomes "Dr.". - Every extra column becomes
{custom.COLUMN_NAME}. OneIndustryorEventcolumn is worth more than ten lines of generic copy. - After
import_contacts, checkrows_invalidandrows_url_inferred. "Inferred" rows carried only a slug: if there are many, the wrong column almost always ended up in the URL field.
Always show the user a preview before enrolling contacts into a campaign.
Writing the first message
It goes in the step after acceptance, never in the invite note (free accounts get 5 notes per month).
Rules that move the acceptance rate more than any setting:
- Spintax is mandatory on the opening:
{Hi|Hello|Hey} {firstName}. Identical messages sent in bulk are one of the strongest bot signals. - One specific reason for writing to this person in particular:
{company},{headline}, or a custom column. If the message works verbatim for anyone, it works for no one. - No pitch in the first message. Its goal is a reply, not a call.
- Stay under 400 characters.
Before launching, show the user the message rendered against 2–3 real contacts from their list, not an invented example.
Reading the numbers
In get_metrics, the only number that really matters is the acceptance rate:
| What you see | What it means | What to do |
|---|---|---|
| > 60% | targeting is right | volume can grow, gradually |
| 40–60% | normal | change nothing |
| < 40% | the list or the message is the problem | the controller brakes on its own: don't fight it, fix the targeting |
entries in signals_7d |
LinkedIn noticed something | lower the volume, don't raise it |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 87 lines · 55 tokens per session scan A 7ac12ac599f7
linkedin-outreach is a skill published in the GitHub repository matteolegrottaglie/curtis (0 stars, last pushed 21d ago), licensed MIT. It adds 55 tokens to every session and 962 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
job-search
Search LinkedIn, Hacker News, and Twitter/X for jobs with connections, hiring manager insights, and preference-based scoring. Use when the user wants to find jobs, search for positions, or explore job opportunities.
job-preferences
Set or update job-search preferences such as titles, salary, remote work, and filters for the other Job Apply skills.
linkedin-safety-verification
Apply security, read-only, authorization, and same-layer verification rules to agent-operated LinkedIn workflows. Use before account-visible changes, private request replay, broad captures, or provider AI tasks.
linkedin-browser-providers
Select and operate the local Playwright, Browserbase, or Anchor Browser provider for an authorized LinkedIn workflow. Use when session persistence, cloud browser attachment, provider profiles/contexts, or explicit hosted browser tasks matter.
linkedin-operator
Operate the user's personal LinkedIn session with the plugin's generic browser and HTTP primitives. Use for authorized LinkedIn reading or changes when the caller must discover the current UI/request shape instead of relying on hardcoded business-action tools.
linkedin-voyager-sdui-methodology
Apply capture-first methodology to LinkedIn Voyager and SDUI traffic without hardcoding mutable endpoints. Use when interpreting captured LinkedIn web requests, query IDs, form state, decorations, or response shapes.