Talent Intel

---
name: talent-intel
description: Build a talent market map for a role or company — talent pools, comp signals, competitor landscape, and sourcing angles — using only free/public web sources. Use when the user asks to map a talent market, scope a role before sourcing, research where candidates for a role come from, or build a market map / talent landscape report.
---

# Talent Intel

Turns a role or company brief into a structured market mapping report. No paid data
sources, no scraping of gated platforms — everything comes from public web search
and public pages.

## When to use this

Trigger on requests like:
- "Map the talent market for a [role] in [location/industry]"
- "Who could we hire for [role] and where do they come from?"
- "Build a market map for [company]'s [function] team"
- "What's the comp landscape for [role]?"

Do **not** use this to research or profile a specific named individual — this skill
maps markets and pools, not people. If the user wants a single-candidate deep dive,
say so and scope down to public, non-invasive research only if they still want it.

## Inputs to collect

Before researching, make sure you have (ask if missing, don't guess silently):

1. **Role** — title and seniority (e.g. "Staff Site Reliability Engineer")
2. **Scope** — industry/domain and geography (e.g. "fintech, SF Bay Area or remote US")
3. **Anchor context** (optional but useful) — a target company, a comparable company,
   or a specific product area, to ground the competitor/pool search

## Flag scope mismatches early

Briefs often anchor on a stage, budget, or comp range that doesn't match where the
relevant talent actually sits right now (e.g. a brief says "Series B-C" but the
live talent pool for that skill set is concentrated at later-stage companies, or a
comp band that's below current market for the role). If research surfaces this
kind of mismatch, say so as soon as it becomes clear — in a short note before or
alongside the report — rather than only burying it in the confidence notes at the
end. The goal is to let the user adjust the brief or their expectations early, not
just read about the gap after the fact.

## Research workflow

Use WebSearch and WebFetch only. Do not log into or scrape LinkedIn, Glassdoor,
levels.fyi, or any gated platform — read only what's publicly indexable, and when a
source requires a login to view details, note it as a source to check manually
rather than working around the gate.

1. **Company/competitor landscape** — identify 5-10 companies that plausibly employ
   people in this role today: direct competitors, adjacent-market companies, and
   companies known for strong teams in this function. Note company size/stage where
   it's public.

2. **Talent pool signals** — for the role, identify recognizable feeder patterns:
   companies people typically move from/to, schools or bootcamps/programs commonly
   associated with the discipline (only where there's real public signal — don't
   invent patterns), and any relevant public communities (open source projects,
   conference circuits, notable public contributors).

3. **Comp signals** — pull public comp ranges where available (levels.fyi public
   pages, published salary bands, public H1B/OFLC wage data, company career pages
   that post ranges). Always cite the source and note it's directional, not
   authoritative — comp data ages fast and varies by leveling.

4. **Sourcing angles** — translate the above into a short list of concrete sourcing
   angles: which companies to target first and why, which public communities to
   watch, and 2-4 **generated** Boolean/LinkedIn X-ray search strings (e.g.
   `site:linkedin.com/in "senior site reliability" (fintech OR payments) -jobs`)
   for the user to run themselves. Never execute these searches on the user's
   behalf against a gated platform — only generate the strings.

## Output

Produce two artifacts:

1. A Markdown report following `templates/report.md.j2`'s section structure.
2. A JSON object conforming to `../../schema/report.schema.json`, so the report can
   feed other tooling.

Keep the report honest about confidence: mark comp numbers and pool-size claims as
estimates, cite sources inline, and flag when public data was thin for a given
section rather than filling gaps with invented specifics.

## Constraints (do not violate)

- No scraping or automated access to LinkedIn or any login-gated platform — search
  strings only.
- No storage or output of named-individual PII — this skill reports on markets and
  pools, not people.
- No paid API keys or subscriptions required to use this skill end to end.