Ken-Technology/cold-email-skills / qualification
qualification
VerifiedWrite AI prompts that judge whether a prospect fits the campaign's audience: who is in, who is out, and why. Use when cleaning or qualifying a prospect list for cold email. Triggers on "qualify my leads", "who should we exclude", "write a qualification prompt".
Qualification Skill
Turn a campaign plan's qualify/disqualify rules into the variables an AI qualifier runs on every prospect: an audience description, disqualification criteria, and (rarely) qualification criteria. They go in qualification.md at the root of the campaign workspace (./cold-email/{slug}/), ready for any qualification step: an AI column in Clay or a spreadsheet, or any LLM.
Inputs
- The plan's broad ICP and its qualify/disqualify rules.
- The search filters already applied (
search-strategy.mdin the campaign workspace). They tell you what NOT to restate. - The research (
research.mdand its company context) for competitors and ICP nuances, and the company knowledge folder (./knowledge/) when it exists: itscompetitors/files and who the ideal customer profile says is not a fit. - A sample of sourced leads, when available, to see real edge cases.
Use what the conversation already holds. Read only what is missing: plan.md, search-strategy.md and research.md in the campaign workspace (default ./cold-email/{slug}/ under the current directory).
What the qualifier can see
The qualifier sees ONLY:
- LinkedIn profile: title, company, location, previous roles, education, skills, summary.
- Company data: industry, headcount, HQ, description.
- Website metadata: page title and meta description only.
It does NOT see full site content, tech stack, news, funding, social posts, press, events, or any time-sensitive signal. Every criterion must be checkable against the data above. Rewrite or drop any rule that depends on something else.
Core Principles
- Write short, trust the AI - Give it a short, plain-language brief and let it reason. A few high-signal lines beat an exhaustive rulebook. No long lists of titles, industries, or product names.
- Never repeat the search filters - The filters already bound the list (headcount, geography, titles, seniority, industries). Do not restate any of it, not in the audience description and not as a criterion. Qualification only catches what the filters can't see.
- Don't over-qualify - Qualification is a safety net, not a precision filter. When data is incomplete or ambiguous, qualify.
- Disqualification over qualification - Prefer audience description + a few disqualifiers. Add qualification criteria only for a real positive signal worth confirming.
- Always exclude competitors - The one disqualifier that is always worth including.
Details, examples, and a disqualifier menu: references/qualification-guide.md.
Write the Variables
Audience description (required): one plain sentence on who we target, 5-20 words. No headcount, geography, title lists, or industries.
Disqualification criteria (strongly preferred): binary deal-breakers the filters can't see, such as competitor or wrong business model. Usually 2-4 lines, never more than 6. Always include competitor exclusion. Don't pad.
Qualification criteria (optional, usually empty): only a genuine positive signal visible in the data above.
At least the audience description and the disqualification criteria must be non-empty. Fewer than two substantive sections is a hard stop: revise first.
Running it
The three sections make one yes/no question per prospect, asked with the company context. Run it where the list lives: an AI column in Clay or a spreadsheet, or any LLM with the prospect's data pasted in. A starting prompt:
You judge whether a prospect fits a cold email campaign. Answer yes or no, then one short reason.
Company: {the company context from research.md}
Audience: {audience description}
Qualifies when: {qualification criteria, or "no extra conditions"}
Disqualified when: {disqualification criteria}
When the data is incomplete or ambiguous, answer yes.
Prospect: {LinkedIn profile, company data, website title and description}
Check 10 to 20 answers by hand before you trust it on the whole list. When you change the criteria, run the list again: earlier answers no longer hold.
Save
Show the three sections to the user and get approval. Then write qualification.md at the root of the campaign workspace:
# Qualification Prompt
## Audience Description
[One plain sentence on who we target. No headcount, geography, titles, or industries.]
## Disqualification Criteria
- [Competitor exclusion]
- [One or two more deal-breakers the filters can't see]
## Qualification Criteria
[Usually empty. Only a real positive signal worth confirming.]
- Exactly these three H2 headings and no others: a tool that reads the file looks for them.
- Keep one qualification file per campaign: when it exists, update its sections, never write a second.
- With no campaign workspace (a one-off request), give the three sections in the chat instead.
Confirm: "Qualification saved to ./cold-email/{slug}/qualification.md."
Recruiting campaigns
A recruiting campaign fills a role: the plan or the user gives the role (title, level, place or remote policy, must-haves). The leads are candidates, so the prompt decides who could take this job, not who would buy.
- Audience description: the role in one or two sentences (title, level, place or remote policy) and the kind of person who could do it.
- Qualification criteria: the role's must-haves, each written as something a profile shows (a current or past title, years of experience, a skill, a degree). Leave out what a profile never shows, such as notice period or salary expectations.
- Disqualification criteria: people who already work at the hiring company, people far more senior or junior than the role, and people whose profile shows none of the must-haves. A nice-to-have never disqualifies.
- Location: for an onsite or hybrid role, people outside the role's area do not qualify unless the role allows relocation. A remote role disqualifies no one on location.
- Never qualify or disqualify on age, gender, race, religion, disability, family status or nationality, or on a guess at any of them (a name, a photo, a graduation year). A language the job needs is a skill, and fine.
Save it exactly as above.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 468k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |