LinkedIn Outbound Skills for AI Agents

Purpose

Find which invite variables drive higher connection acceptance so you stop guessing at copy and targeting.

Inputs

• Export of sent connection requests with status (accepted, pending, ignored)
• Invite note text used per request
• Profile fields per recipient: title, seniority, company size, industry

Steps

1. Load the sent request export and normalize status into three buckets: accepted, pending, ignored.
2. Tag each request with the invite variant used (note vs no note, hook type, length).
3. Cross each variant against recipient traits and compute acceptance rate per combination.
4. Flag any cell with fewer than 20 requests as low confidence so you do not over read noise.
5. Rank combinations by acceptance rate and list the bottom five that waste invite quota.
6. Write two or three plain observations, such as which seniority ignores noted invites.

Output

A table of acceptance rates by variant and trait, a low confidence flag column, and a short list of what to keep sending and what to cut.

Purpose

Turn a prospect profile into three tested connection request notes and pick the one most likely to get accepted.

Inputs

• Prospect name, title, and company
• Recent activity or a post they made
• Your reason for reaching out in one line

Steps

1. Pull the profile details and any recent post or comment into a short summary.
2. Identify one specific detail that shows you read their profile, not a generic role reference.
3. Draft three note variants under 300 characters each, each leading with the specific detail.
4. Score each variant on relevance, brevity, and whether it asks for nothing in the first message.
5. Flag any note that pitches, sells, or uses filler like 'love your content'.
6. Rewrite the top variant to remove the weakest phrase.

Output

Three ranked connection notes with scores and a one line reason the top pick wins.

Purpose

Warm a prospect by leaving three useful comments on their recent posts before any outreach so the connection request lands with recognition.

Inputs

• Prospect profile URL
• Their last 5 to 10 posts (text pasted in)
• Your area of expertise and point of view

Steps

1. Read each post and tag it by topic, stance, and how much engagement it already has.
2. Pick the 3 posts where you can add a specific insight, not just agreement.
3. For each, draft a comment that references one detail from their post and adds one concrete example or number.
4. Keep each comment under 40 words and remove any pitch or link.
5. Space the comments across 3 different days and note the suggested order.
6. Write a connection note that references the comment thread you built.

Output

A table of 3 dated comments, the post each targets, and a matching connection note referencing the shared thread.

Purpose

Convert a shared event into a natural connection request and follow-up DM that does not pitch.

Inputs

• Event or webinar name and date
• Prospect name, title, company
• Your reason for attending or one takeaway you had

Steps

1. Confirm the prospect actually attended or engaged with the event; skip anyone you cannot verify.
2. Write a connection note under 200 characters that names the specific event and one shared point of interest, with no ask.
3. Draft a first DM to send two to three days after they accept, referencing your takeaway and asking one open question about their view.
4. Add a soft second DM for one week later that offers a relevant resource, not a meeting.
5. Flag any prospect whose profile shows they already bought a competing tool so you can adjust the angle.
6. Keep all messages in plain sentences, no bullet dumps, no links in the first two touches.

Output

A connection note plus a two-message DM sequence with send-day timing for each prospect.

Purpose

Spread outreach across days and hours so activity looks human and stays under safe limits.

Inputs

• Total prospect count and list source quality
• Account age and current weekly connection acceptance rate
• Your working days and time zone

Steps

1. Set a daily connection cap based on account age, starting low for newer accounts and ramping weekly.
2. Split the cap into two or three sending windows per day rather than one burst.
3. Reserve a portion of daily actions for replies and profile visits so the account is not only sending requests.
4. Insert one to two rest days or lighter days each week to break the pattern.
5. Add randomized minute offsets to each window so sends do not land on the same clock time.
6. Recalculate the cap down if acceptance rate drops below a set floor, since that signals over-sending.

Output

A day-by-day and window-by-window sending schedule with per-slot action counts and ramp notes.

Purpose

Turn a loose ideal customer description into a tight, reproducible Sales Navigator search.

Inputs

• Your ICP in plain words, including industry, size, and buying trigger
• Titles of people who typically buy and who blocks the deal
• Any regions or company types to exclude

Steps

1. Map the ICP to concrete Sales Navigator fields: headcount bands, industries, seniority, and function.
2. Build a primary title list plus close variants, since one title rarely covers a role.
3. Add exclusion filters for wrong seniority, agencies, students, and competitors.
4. Layer a trigger filter such as recent job change, hiring, or posting keyword to sharpen intent.
5. Estimate the resulting list size and tighten or loosen one filter at a time until it lands in a workable range.
6. Write down the exact filter set so the search can be rebuilt or handed off later.

Output

A documented filter stack with primary filters, exclusions, a trigger layer, and the expected list size.

Purpose

Surface the best shared connection between you and a target account contact, then draft a low-friction introduction request.

Inputs

• Target contact profile URL and company
• Your own connection list export or Sales Navigator access
• Your one line offer and why this target fits

Steps

1. Pull the list of shared connections between you and the target contact.
2. Score each shared connection on relationship strength using recency of interaction, shared employer history, and title relevance.
3. Rank the top three paths and note why each person can vouch credibly.
4. Draft a short intro-ask message to the highest ranked connection that states the target, the reason, and an easy opt-out.
5. Draft a forwardable blurb the connection can paste directly to the target.
6. Flag any path where the relationship looks too weak to ask, and suggest a direct approach instead.

Output

A ranked list of warm paths with strength notes, one intro-ask message, and one forwardable blurb ready to send.

Purpose

Turn reactions and comments on a target post into warm, context-rich outreach.

Inputs

• The post URL and a summary of its topic
• A list of engagers with name, title, and their comment if any
• Your offer in one plain sentence

Steps

1. Sort engagers into commenters and passive reactors, since commenters get a more specific opener.
2. For each commenter, quote or paraphrase their actual comment in the connection note to prove you read it.
3. For reactors, reference the post theme and ask what pulled them to it.
4. Write a first DM that ties the post topic to a problem your offer addresses, ending in a question, not a calendar link.
5. Add a break-up DM for ten days later that closes the loop politely and leaves the door open.
6. Cap total touches at three and space them at least three days apart.

Output

Per-engager connection notes and a three-touch DM sequence, grouped by commenter versus reactor.

Purpose

Determine the days and hours a target prospect is most active on LinkedIn so outreach lands when they are likely to read it.

Inputs

• Prospect profile URL
• Recent post and comment timestamps (last 30 days)
• Prospect timezone or headquarters location

Steps

1. Pull the prospect's last 15 to 20 public activities including posts, comments, and reactions with their timestamps.
2. Convert every timestamp to the prospect's local timezone using their listed location.
3. Bucket the activity into weekday and hour-of-day groups to spot repeated windows.
4. Rank the top two windows by frequency, breaking ties toward mid-week mornings.
5. Flag whether the account is dormant (no activity in 14 days) and downgrade priority if so.
6. Draft a one-line send recommendation naming the exact day and hour range to fire the connection request or message.

Output

A short table per prospect with top active windows, dormancy flag, and a recommended send time in their local timezone.

Purpose

Convert people who viewed your LinkedIn profile into a lightweight DM sequence that opens conversation without a hard pitch.

Inputs

• List of recent profile viewers with name, title, company
• Your ICP definition and offer angle
• Your recent post topics or content themes

Steps

1. Filter the viewer list down to accounts that match the ICP and drop anyone already in an active sequence.
2. Tag each viewer by likely intent, either content-driven, hiring-driven, or competitor curiosity, based on their title and activity.
3. Draft a first DM that references the shared context of the view without stating you saw them view you.
4. Write a second message for three days later that offers one specific resource tied to their role.
5. Write a third message for day seven that asks a single low-pressure question to invite a reply.
6. Set a stop rule so any reply or connection removes the person from the remaining steps.

Output

A per-viewer table with intent tag, three drafted DMs, send day offsets, and stop conditions ready to load into a manual send queue.

Purpose

Convert a written LinkedIn message into a short voice note script that sounds spoken, not read.

Inputs

• Target profile name, role, and company
• The reason you are reaching out (trigger or shared context)
• Your one line offer or ask
• Preferred length in seconds, 20 to 40

Steps

1. Pull the strongest specific detail from the profile or trigger to open with.
2. Draft the script in spoken cadence with short sentences and natural filler removed.
3. Front load the name and the reason within the first eight seconds.
4. State one clear ask near the end, phrased as a low pressure question.
5. Mark pause points and words to stress so delivery stays relaxed.
6. Trim until read time matches the target second count at normal pace.
7. Add a fallback text line to paste under the note for skimmers.

Output

A timed voice note script with stress marks, a spoken word count, and a paired one line text caption.