agentleFS
vivekkhimanivivekkhimani/gtm-tools-template/.agents/skills/gtm-contact-filter/SKILL.md

gtm-pipeline:contact-filter

Skill1 starsChanged 5 months agoMITView on GitHub

Classify and filter contacts from people-search output BEFORE enrichment. Rejects non-ICP contacts and ranks the rest using job tier, industry tier, location tier, company size, and ICP keyword scoring. Saves enrichment credits by only passing qualified contacts downstream. Input: contacts_found.csv. Output: contacts_filtered.csv. Also triggers on "filter contacts", "ICP filter", "rank contacts".

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 3 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    After people-search, before people-enrichment

    line 16
  2. Very specific:

    Persona workflow: Run PB Profile Scraper with enrichWithCompanyData: true during/after people-search.

    line 38
  3. Distinctive:

    Standalone: Run one of the API providers above before contact-filter.

    line 39
  4. Very specific:

    Read {client-slug}-gtm/context/icp.md for client-specific tier definitions, thresholds, and sort order.

    line 49
  5. Very specific:

    C-level promotion rule: If company_employee_count < {client_clevel_threshold} (e.g.

    line 80
  6. Very specific:

    Prefer not to reject solely on missing data — use the hard_reject_threshold to control strictness per client.

    line 120
  7. Very specific:

    Ask the user at setup which ordering fits their priorities.

    line 143
  8. Very specific:

    Default policy: Don't hard-reject on missing data unless the corresponding tier is explicitly above threshold.

    line 179
  9. Very specific:

    Run the classification logic over all rows in contacts_found.csv.

    line 189
  10. Distinctive:

    All input columns preserved.

    line 217

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Contact Filter

Classify and rank contacts from people-search output using a three-tier ICP scoring system. Rejects non-ICP contacts before enrichment to save credits.

Read .agents/skills/_shared/conventions.md before executing.


When to Use

  • After people-search, before people-enrichment
  • Input: csv/intermediate/contacts_found.csv
  • Output: csv/intermediate/contacts_filtered.csv
  • Even small batches (10 contacts) benefit from ICP ranking

Prerequisites — Company Data

Contact-filter classifies on company-level fields (company_employee_count, company_hq_location, company_industry, company_specialities) alongside each contact. These come from upstream skills.

SourceFields ProvidedCost
PB LinkedIn Profile Scraper (enrichWithCompanyData: true)employee_count, hq, industry, specialitiesFree (PB plan + LinkedIn account)
PB Sales Navigator Account Scraperemployee_count, hq, industry, growth metricsFree (PB plan + SN account)
PB LinkedIn Company ScraperFull company page dataFree (PB plan + LinkedIn account)
FullEnrich Company Finderemployee_count, hq, industry0.25 credits/company
Pipe0 companies:profiles:amplemarket@1employee_count, hq, industry2.00 credits/page (100 results)
Pipe0 companies:profiles:crustdata@1employee_count, hq, industry2.00 credits/page (100 results)

Where company data typically comes from:

  • Company-First workflow: Already enriched in the company-enrichment step.
  • Persona workflow: Run PB Profile Scraper with enrichWithCompanyData: true during/after people-search.
  • Standalone: Run one of the API providers above before contact-filter.

If company data is missing: The filter falls back to job-title-only classification. It skips company-dependent dimensions (industry tier, location tier from HQ, size filter) and relies on job_tier and person location only.

See _shared/conventions.md for the canonical field name mapping (snake_case across all skills).


Step 1 — Load ICP Configuration

Read {client-slug}-gtm/context/icp.md for client-specific tier definitions, thresholds, and sort order.

If icp.md does not yet contain contact-filter configuration, ask the user:

  1. What are your target job titles / role keywords?
  2. What seniority levels are you targeting? (manager, director, VP, C-level?)
  3. What industries are highest priority? Lowest priority?
  4. What geographies are highest priority?
  5. What company size range? (min and max employee count)
  6. Any ICP keyword signals to look for in company specialities?
  7. Which dimensions matter most for ranking? (e.g. location first, then role, then industry)

Save the answers to context/icp.md under a ## Contact Filter section.


Classification System

A. Job Tier (1–6, lower = better)

Fully parameterized — client defines role keywords. Example structure:

TierDescriptionExample
1Primary target roles, mid-level seniorityRevOps Manager, Sales Ops Lead
2Primary target roles, other seniorityRevOps IC, Head of RevOps
3Leadership in target domainVP Sales, Director of Revenue
4General domain rolesAccount Executive, Sales Manager
5C-level (promote to tier 3 if company < threshold employees)CRO, CSO
6Reject (non-target roles, junior, students)Intern, Coordinator, Marketing Analyst

C-level promotion rule: If company_employee_count < {client_clevel_threshold} (e.g. 50), promote C-level from tier 5 → tier 3 (they're operational at small companies).

Hard reject: job_tier >= hard_reject_threshold (default: 6).

Missing title: tier 6 (Unknown). Default threshold rejects tier 6. To allow missing-title contacts through, set hard_reject_threshold: 7 in your ICP config.

B. Industry Tier (0–5, lower = better)

Client defines 5 industry tiers ranked by fit for their offering.

  • Match logic: exact match first (full industry name), then keyword fallback
  • Tier 0 = Unknown (company_industry missing) — sorted between tier 1 and tier 2
  • Missing industry → tier 0, NOT rejected

C. Location Tier (A–D, A = best)

Client defines geographic priority tiers.

  • Primary: company HQ location (company_hq_location field — format: "City, CC", extract country code CC)
  • Fallback: person location field
  • Both missing: tier D (lowest)
  • Hard reject option: client can specify only tier A (or A+B) passes

D. Company Size Filter

Client defines min and max company_employee_count.

  • Outside range AND field is present → hard reject
  • Field missing/empty → skip size filter, set employee_count_missing: true, deprioritize in sort

E. ICP Keyword Scoring

Scan company_specialities for client-defined domain keywords. Count matches → icp_kw_score (0 = no match, no penalty). Missing specialities → icp_kw_score = 0.


Hard Filters (Rejection Rules)

Prefer not to reject solely on missing data — use the hard_reject_threshold to control strictness per client. The exception is missing job title (tier 6), which is rejected at the default threshold; raise the threshold to 7 to keep them.

FilterConditionAction
Job tier>= hard_reject_threshold (default: 6)Reject
Employee countoutside [min, max] AND field presentReject
Location tierworse than threshold AND field presentReject (optional)

Soft Sort (Ranking)

Sort order is configurable per client — stored in context/icp.md.

Default composite sort key (used if no preference stated):

job_tier → industry_tier → company_loc_tier → person_loc_tier → employee_count_missing (penalty last) → icp_kw_score (desc) → follower_count (desc)

Other valid orderings:

  • Geography-first: company_loc_tier → job_tier → industry_tier → ...
  • Company-fit-first: industry_tier → icp_kw_score → company_loc_tier → job_tier → ...

Ask the user at setup which ordering fits their priorities. Store it in context/icp.md.


Output Columns Added

The following columns are appended to all passed records:

ColumnDescription
job_tier1–6 (1 = best fit)
personaLabel for the role category
seniorityDetected seniority level
ind_tier0–5 (0 = unknown, 1 = best fit)
ind_labelMatched industry label
comp_loc_tierA–D based on company HQ
person_loc_tierA–D based on person location
effective_loc_tierBest of comp_loc_tier / person_loc_tier
employee_count_missingtrue/false
icp_kw_scoreCount of ICP keyword matches in specialities
priority_labelHuman-readable priority (e.g. "High", "Medium", "Low")
sort_keyComposite sort value used for ranking

Missing Data Handling

FieldMissing Behavior
job_titletier 6 (Unknown) — rejected at default threshold (6); raise to 7 to keep
company_industrytier 0 (Unknown) — sorted between tier 1 and tier 2
company_hq_locationFall back to person location; if both missing → tier D
company_employee_countSkip size filter, flag employee_count_missing: true, deprioritize in sort
company_specialitiesicp_kw_score = 0 (no penalty)
Follower count0 (lowest tiebreaker priority)

Default policy: Don't hard-reject on missing data unless the corresponding tier is explicitly above threshold. Adjust hard_reject_threshold per client.


Execution Protocol

1. Read ICP Configuration

Load from context/icp.md. If contact-filter section is missing, ask the setup questions above and save the answers.

2. Classify All Contacts

Run the classification logic over all rows in contacts_found.csv. Add the output columns listed above.

3. Apply Hard Filters

Reject rows that fail hard filter conditions. Log rejection reasons.

4. Sort Passed Records

Apply the client's configured sort order (or default) to rank passed contacts.

5. Print Run Summary

## Contact Filter Summary
- Input: {n_input} contacts
- Hard rejected (job tier): {n_job_rejected}
- Hard rejected (employee count): {n_size_rejected}
- Hard rejected (location): {n_loc_rejected}
- Passed: {n_passed} contacts

## Tier Distribution (passed contacts)
Job tiers: tier1={x}, tier2={y}, tier3={z}, tier4={w}
Industry tiers: tier1={x}, tier2={y}, tier3={z}, ...
Location tiers: A={x}, B={y}, C={z}, D={w}

## Top 10 by Priority
[name | company | job_title | job_tier | ind_tier | loc_tier | icp_kw_score]

6. Write Output

Save passed contacts (ranked) to csv/intermediate/contacts_filtered.csv. All input columns preserved. Output columns appended.


ICP Configuration Format (in context/icp.md)

## Contact Filter

### Job Tier Definitions
- Tier 1 (Primary mid-level): keywords = [revops, sales ops, revenue operations], seniority = [manager, senior, lead]
- Tier 2 (Primary other): keywords = [revops, sales ops], seniority = [head of, director, IC]
- Tier 3 (Leadership): keywords = [sales, revenue, commercial], seniority = [VP, director, head of]
- Tier 4 (Domain): keywords = [account, sales, business dev], seniority = any
- Tier 5 (C-level): keywords = [CRO, CSO, chief revenue, chief sales]
- Hard reject (Tier 6): keywords = [student, intern, junior, coordinator]
- C-level promotion threshold: 50 employees
- Hard reject threshold: tier >= 6

### Industry Tiers
- Tier 1 (Best fit): [B2B services, manufacturing, logistics, wholesale, consulting]
- Tier 2 (Good fit): [retail, hospitality, consumer-facing with B2B sales]
- Tier 3 (Neutral): [SaaS, software, IT services]
- Tier 4 (Low fit): [deeptech, aerospace, biotech]
- Tier 5 (Hard): [finance, healthcare, pharma, legal, government]

### Location Tiers
- Tier A: [DE, AT, CH, NL, BE, LU, DK, SE, NO, FI, GB, IE]
- Tier B: [FR, ES, IT, PL, CZ, HU, RO, PT, GR, BG, HR, SK, SI]
- Tier C: [US, CA, AU, NZ]
- Tier D: all other
- Hard reject threshold: none (all tiers pass, used for sorting only)

### Company Size Filter
- Min: 10 employees
- Max: 500 employees

### ICP Keywords (company specialities)
- keywords = [automation, workflow, crm, sales enablement, revops, salestech]

### Sort Order
job_tier → industry_tier → comp_loc_tier → person_loc_tier → employee_count_missing → icp_kw_score (desc)

Reference Patterns

The classification logic follows these patterns:

  • classify_job(title, employee_count) → returns (tier, persona, seniority) based on keyword matching + seniority detection + C-level promotion rule
  • Industry: exact set match first, keyword fallback
  • Location: parse country code from "City, CC" format in HQ field
  • Composite sort key as tuple: (job_tier, ind_tier, loc_tier_ord, ...) — Python's natural tuple ordering does the ranking

What's Missing (To Document)

  • Benchmark: credit savings (% reduction) from filtering before enrichment across different client ICPs
  • Native integration with downstream tooling (n8n, custom scripts)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.