Okhp3 Linkedin Message
ASecurityDraft a LinkedIn direct message or a counter reply to an inbound message from raw dictation, notes, or the received text. Use when the user wants to write a DM, reconnect after a long gap, answer a new connection's message, follow up after an accepted invite, reply to a recruiter or peer, or coordinate several related messages about one subject. It tiers the recipient by role and relationship temperature, verifies recipient facts, matches length, keeps one small ask, scrubs risk, and returns ...
- 3 stars
- 0 votes
- 0 copies
- 0 views
- Added November 18, 2026
Security analysis
100/100Pro scans all 6 files and shows the line behind each finding
npx -y skills add OKHP3/skillz --skill okhp3-linkedin-message --agent claude-codeAre you the author of Okhp3 Linkedin Message?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/okhp3-okhp3-linkedin-message)---
name: okhp3-linkedin-message
description: >
Draft a LinkedIn direct message or a counter reply to an inbound message from
raw dictation, notes, or the received text. Use when the user wants to write
a DM, reconnect after a long gap, answer a new connection's message, follow
up after an accepted invite, reply to a recruiter or peer, or coordinate
several related messages about one subject. It tiers the recipient by role
and relationship temperature, verifies recipient facts, matches length,
keeps one small ask, scrubs risk, and returns drafts only. Do not use for
comments, posts, connection notes, or sending.
license: MIT
metadata:
public_artifact: true
author: Public Agent Skill Library
version: "1.1.0"
category: social-posting
origin: portable-agent-skills
in_scope: "Drafting LinkedIn direct messages or counter replies from supplied sender intent and recipient context, including related messages with a send order, with a risk scrub."
out_of_scope: "Sending, comment or post drafting, connection notes, bulk messaging, or inventing facts about either party."
---
# okhp3-linkedin-message
Portable, identity-agnostic Agent Skill.
## Public-package data boundary
This package and its fixtures contain no execution account, person, project,
campaign, topic, bias, opinion, or private URL. Use placeholders in public
examples and obtain all such context from the current run only.
A good message is short enough to answer on a phone and specific enough to be
worth answering.
## Scope
This package drafts a direct message or counter reply, or a small set of
related messages about one subject with a send order. It does not send,
batch, or schedule messages. For first-touch invite text use
`okhp3-linkedin-connection-note`. For public conversation use
`okhp3-linkedin-comment`.
## Process
1. Gather inputs: the sender's raw intent in their own words, the inbound
message when replying, whatever the sender supplied about the recipient, and
any standing constraints the sender wants honored. A dictated thought is
source material, not a finished draft. Do not invent history between the
parties.
2. Tier the recipient using supplied evidence only: peer, senior or high-reach
contact, recruiter or sales sender, or unknown. The tier sets tone and
length. Senior contacts get shorter and more specific. Recruiter and sales
senders get a verification-minded reply, not instant enthusiasm. Unknown
senders get no personal data.
Then set relationship temperature with `references/relationship-tiers.md`:
`active`, `dormant`, `distant`, or `cold`. Temperature sets candor,
attachments, and how the gap or lack of history is acknowledged. Do not
infer warmth the sender did not state.
3. Verify recipient facts before using them. A statement about the
recipient's current role, employer, or recent change needs a source from
this run, preferably the recipient's own current profile. Flag a press
release or third-party page as `needs-confirmation`; it may be stale.
Describe a role change neutrally.
4. Run the recipient-conflict check. If the recipient works for a party the
message or an attachment treats as a competitor, counterparty, or target,
remove that material or return `needs-decision`. Attachments follow
temperature: `active` may carry working files; `dormant` and `cold` carry
none in the first message; `distant` carries at most one that previews
inline when the host supports it.
5. Match length to the inbound. A two-line message gets a two- or three-line
reply. A long thoughtful message earns a proportionate answer. Cap first-
touch messages at a few short sentences.
6. State the single purpose in one sentence before drafting. Include at most
one small, easy-to-grant ask, such as a viewpoint, a resource pointer, or a
short call later. Do not stack asks. Do not pitch in a first message.
7. Draft in the sender's voice by applying `okhp3-linkedin-voice`, which
covers contractions, no em dashes, and preserved uncertainty. Keep the
sender's real phrasing where it works. Do not sand it into corporate paste.
8. Run `references/message-gate.md`. Scrub employer-identifying, confidential,
or agreement-restricted content and any statement about employment status
the sender did not supply. Do not include the sender's phone, email, or
private links unless the sender asked.
9. When several messages concern one subject, order them with
`references/sequencing.md` (ally, then target, then mutuals who might tell
the target), make every cross-reference true on arrival, and read each
draft as if it will be forwarded to every other recipient.
10. Return the draft, a one-line rationale, and the gate report. If the
inbound message looks like a scam, phishing, or mass-mail pitch, say so and offer a
verification reply instead of an engaged one.
11. Do not send. Show the final text for approval. Sending is a separate host
action that needs explicit approval for that message.
## Output contract
Return `Draft`, `Why this shape` in one line, and `Gate report` naming the
role tier, temperature, sourced recipient facts, conflict-check result, and
attachments. For related messages, add `Send order` with the condition each
step waits on. When the sender intent is unclear in a way that changes the message, ask at most one
question. Otherwise assume, label the assumption, and draft.
## Validation
Before return, verify the length fits the tier, there is one ask or none, no
unsupplied fact appears, the voice gates ran, no sensitive contact data leaked,
and no sending action is included.
## Platform boundary
LinkedIn can restrict messaging volume, limit messages to connections, and
flag automated or repetitive messaging. A draft does not establish that the
message can be sent or that the recipient allows it.
## Public-neutrality gate
Keep this package portable for public distribution. Do not embed a person name,
account handle or URL, private identifier, employer or organization identity,
product or project name, campaign name, or user-specific subject in instructions,
examples, references, or evaluation fixtures. Resolve account, destination,
audience, subject, links, current facts, and visibility from execution-time
context. Use placeholders or supplied evidence; platform help links may remain.
## About
This package is distributed as a portable, identity-agnostic Agent Skill.
It resolves account, destination, audience, subject, and source context at execution time.
MIT License -- free to use, fork, and adapt. A nod to the source is appreciated.
Files in this skill
- SKILL.md
- benchmarks/equilibrium-review-2026-11-09.json
- evals/evals.json
- references/message-gate.md
- references/relationship-tiers.md
- references/sequencing.md
Attribution
Related skills
View all in ai-agents →Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.