The first 7 of 13 new skills, inspired by real patterns found in xx254/linkedin_skills and mcpmarket's LinkedIn Lead Finder, written as fully original skills.
Skill 42: lead-scoring-tiers
Inspired by the qualified/disqualified CSV pattern seen in xx254/linkedin_skills, a full lead-filtering skill that outputs a canonical registry rather than a one-off score.
mkdir -p .claude/skills/lead-scoring-tiers
---
name: lead-scoring-tiers
description: >
When scoring a CSV of LinkedIn leads into qualification tiers, load
this skill.
---
Input: leads.csv (name, title, company, LinkedIn URL). ICP from CLAUDE.md.
For each lead, score against: title match, company type match, size
fit, geography, and any disqualifier present.
Bucket into: qualified.csv, review.csv (borderline, needs a human
judgement call), disqualified.csv (with the specific reason logged).
Also write lead_registry.json: a canonical, deduplicated record of
every lead ever scored, so re-runs never re-score or re-contact the
same lead twice.
Rules: never silently drop a lead, every lead ends up in exactly one
of the three output files with a reason logged.
Skill 43: voice-calibration
Inspired by the /calibrate-voice pattern seen in xx254/linkedin_skills, a skill that builds and continuously updates a brand-voice.md file from real examples rather than a one-time description.
mkdir -p .claude/skills/voice-calibration
---
name: voice-calibration
description: >
When building or updating a writing voice guide from real message
examples, load this skill.
---
Input: 5-10 real messages, posts, or comments the user has actually
written and considers on-voice.
Analyse: sentence length pattern, vocabulary choices, how directly they
state things, what they never say, how they open and close messages.
Write or update brand-voice.md with these findings as concrete rules,
not vague adjectives.
On subsequent runs: if given new examples plus feedback on what worked
or did not, update brand-voice.md with the specific correction rather
than rewriting the whole file.
Skill 44: decision-maker-mapper
Inspired by mcpmarket's LinkedIn Lead Finder pattern of mapping job titles to departmental functions with confidence scoring.
mkdir -p .claude/skills/decision-maker-mapper
---
name: decision-maker-mapper
description: >
When identifying the right decision-maker or department head at a
target company, load this skill.
---
Input: target company, the function I need a contact for (e.g.
Marketing, Engineering, Ops).
Map the function to the specific titles likely to hold decision-making
authority at a company of this size and type (a 20-person company's
Head of Marketing differs from a 2,000-person company's VP Marketing
in actual authority level).
Search for matching profiles. For each candidate: confidence score
(high/medium/low) based on title match, tenure, and any visible signal
of actual authority (do they post about budget decisions, do they
appear in the company's own announcements).
Output: ranked candidate list with confidence scores, not a single
guess presented as certain.
Skill 45: campaign-diagnostic
Inspired by the diagnostic pattern in xx254/linkedin_skills that identifies which lever (targeting, voice, or reply handling) is actually broken when a campaign underperforms.
mkdir -p .claude/skills/campaign-diagnostic
---
name: campaign-diagnostic
description: >
When a LinkedIn outreach campaign is underperforming and needs
diagnosis, load this skill.
---
Input: campaign metrics (connections sent, accepted, DMs sent, replies,
meetings booked). Sample of actual messages sent.
Diagnose which lever is broken:
- Targeting: low acceptance rate despite a decent connection note
suggests the ICP itself is off, not the message
- Voice: decent acceptance but low reply rate suggests the DM content
or tone is the problem
- Reply handling: good reply rate but few meetings booked suggests the
conversation itself is not converting
Output: a precise diagnosis (not a vague "try harder"), a specific fix
list, and if the fix involves voice, a note to run voice-calibration
with corrected examples.
Skill 46: amplify-connector-setup
Inspired by ConnectSafely's Amplify Claude connector pattern, describing how to combine persistent memory, an API connection, and sub-agents into one setup, written as an original configuration guide skill rather than the specific vendor implementation.
mkdir -p .claude/skills/amplify-connector-setup
---
name: amplify-connector-setup
description: >
When setting up a persistent, memory-backed LinkedIn automation
connector, load this skill to check the configuration is complete.
---
Check for 3 components before considering the setup complete:
1. Persistent memory: a location where campaign history, contacted
leads, and voice examples accumulate across sessions rather than
resetting each time
2. The API connection: whichever LinkedIn automation API or MCP is
being used, confirmed connected and authenticated
3. Sub-agent delegation: whether specific tasks (research, drafting,
reply handling) are split across dedicated skills/agents rather
than one general-purpose prompt handling everything
Output: a checklist confirming all 3 are in place, or specific gaps
to fix before the system is considered production-ready.
Skill 47: reciprocity-tracker
mkdir -p .claude/skills/reciprocity-tracker
---
name: reciprocity-tracker
description: >
When logging daily comment activity to avoid repeating the same
creators or prospects in a way that looks automated, load this skill.
---
Log every comment left: which creator or prospect's post, the date,
and a brief note on the angle taken.
Before each new comment session: check the log so the same person is
not commented on twice within a short window, and so comment angles
are varied rather than repeating the same observation.
Output: a running reciprocity-log.md and a flag if any single
creator or prospect is being over-targeted relative to the rest of
the list.
1:1 with a GTM engineer
