| name | tenacious-signal-brief |
| description | Generate a hiring signal brief and competitor gap brief for a Tenacious Consulting prospect from public data. Use this skill whenever the user mentions building outbound for Tenacious, enriching a prospect, the "hiring signal brief," "competitor gap brief," scoring AI maturity (0-3), combining Crunchbase + layoffs.fyi + job-post + leadership signals, or any phrasing like "research this company before we email them." Always use this instead of ad-hoc enrichment — the honesty and confidence-language rules (no "aggressive hiring" below five open roles, confidence-weighted phrasing, no fabricated numbers) are Tenacious brand constraints, not style preferences, and getting them wrong damages the client's reputation. |
Tenacious Hiring Signal Brief
A prospect-enrichment skill that produces two JSON artifacts before any outbound:
hiring_signal_brief.json — a per-prospect view of the five signals that define their buying window.competitor_gap_brief.json — the prospect's AI-maturity position against 5–10 top-quartile peers in their sector.
Together these convert outbound from "you might need offshore engineering" into a grounded research finding the prospect cannot object to without contradicting their own public record.
Why the honesty rule is load-bearing
Tenacious's reputation is the moat. Over-claiming a signal ("you are scaling aggressively" when the prospect has three open roles) costs more than silence. Every field in the output carries a confidence score, and the agent that consumes this brief MUST shift its language when confidence is low — asking rather than asserting. If you cannot ground a claim in the brief, leave the field null. A null field is a signal to the downstream agent to omit that line, not to improvise.
Input
Either:
- A company name + domain, OR
- A Crunchbase ODM record ID
Seed data available (from the challenge seed repo):
- Crunchbase ODM sample (1,001 companies, Apache 2.0)
- layoffs.fyi CSV (CC-BY)
- Public job-post snapshot (early April 2026) or small live crawl (≤200 companies, no login, respect robots.txt)
The five signals
| Signal | Source | Output field |
|---|
| Funding event (Series A/B in last 180 days) | Crunchbase ODM + press | funding.event, funding.amount_usd, funding.closed_at, funding.confidence |
| Job-post velocity (60-day delta) | BuiltIn, Wellfound, LinkedIn careers pages | hiring.open_eng_roles, hiring.delta_60d, hiring.confidence |
| Layoffs (last 120 days) | layoffs.fyi | layoffs.event, layoffs.headcount_pct, layoffs.closed_at, layoffs.confidence |
| Leadership change (CTO/VP Eng in last 90 days) | Crunchbase + press | leadership.change, leadership.role, leadership.appointed_at, leadership.confidence |
| AI maturity score (0–3) | Job posts + team page + GitHub org + exec talks + tech stack | ai_maturity.score, ai_maturity.inputs[], ai_maturity.confidence |
AI maturity scoring — the part most agents get wrong
The score is a 0–3 integer with per-input justification. Weights:
- High: AI-adjacent open roles as a fraction of total eng openings; named AI/ML leadership on public team page.
- Medium: Public GitHub org commits on model/inference repos; CEO/CTO posts or keynotes in last 12 months naming AI as strategic.
- Low: Modern data/ML stack signal (dbt, Snowflake, Databricks, W&B, Ray, vLLM via BuiltWith/Wappalyzer); annual-report or fundraising-press AI positioning.
Rules:
- Absence is not proof of absence. Many companies keep AI work private. A score of 0 with low input count is "no public signal," not "no AI work."
- Score 2 from weak inputs ≠ score 2 from strong inputs. Emit a
confidence field (low/medium/high) alongside the score.
- A readiness of 2 inferred from one medium-weight input must be flagged low-confidence so the downstream agent softens the language.
Competitor gap brief
Once AI maturity is scored:
- Identify 5–10 top-quartile competitors in the prospect's sector and size band using Crunchbase industry + employee-count filters.
- Run the same AI-maturity scoring on each.
- Compute the prospect's position in the sector's distribution (percentile + rank).
- Extract 2–3 specific public-signal practices the top quartile shows that the prospect does not (e.g., "three of five top-quartile peers have a Head of AI on the public team page; prospect does not").
The output is a research finding, not a pitch. The downstream agent uses it to lead with a gap observation, then pivots to Tenacious capability only if the prospect engages.
How AI maturity changes the pitch (downstream consumers care)
Emit a pitch_guidance field the email drafter reads:
- Segment 4 (capability gap / ML platform / agentic): Only viable at
ai_maturity.score >= 2. Below that, flag segment_4_viable: false so the email drafter does not reach for it.
- Segment 1 (recently funded): Always viable, but language shifts. High readiness → "scale your AI team faster than in-house hiring supports." Low readiness → "stand up your first AI function with a dedicated squad."
- Segment 2 (mid-market restructuring): Same shift as Segment 1.
- Segment 3 (leadership transition): Unaffected — the new leader's stance is the variable, not prior state.
Output schema
Emit two files. Exact schema:
{
"prospect": {"company": "string", "domain": "string", "crunchbase_id": "string"},
"enriched_at": "ISO-8601 timestamp",
"funding": {
"event": "Series A | Series B | null",
"amount_usd": 14000000,
"closed_at": "2026-02-14",
"confidence": "high | medium | low",
"sources": ["url1", "url2"]
},
"hiring": {
"open_eng_roles": 27,
{
"prospect": {"company": "string", "sector": "string", "size_band": "50-200"},
"cohort": [
{"company": "string", "ai_maturity": 3, "source_urls": ["..."]}
],
"prospect_position": {"percentile": 35, "rank": "7 of 10"},
"gaps": [
{
"practice": "Named Head of AI on public team page",
Output discipline
- Null over guess. A field you cannot ground goes null with no fallback text.
- Every non-null field carries a
confidence value and a sources array. No exceptions. The email drafter refuses to quote uncited claims.
- No fabricated numbers. Making up a funding amount or headcount is a disqualifying violation per the evidence-graph grading rule.
- Stamp
enriched_at. HubSpot records require current enrichment timestamps; stale briefs are a tracked failure mode.
Probe-triggering failure modes to avoid
These are the exact adversarial probes the challenge will grade you on. Fail them here and you fail them downstream:
- ICP misclassification. A company with both a recent layoff AND recent funding goes into Segment 2 (restructuring), not Segment 1 (freshly funded). Flag both signals; let the ICP classifier decide.
- Signal over-claiming. Fewer than 5 open roles → do NOT emit
hiring.delta_60d as "aggressive." Emit the raw number; let the downstream agent decide the language.
- AI-maturity false positive. A "loud but shallow" company (one exec keynote, no team, no roles) must not score 3. Require at least one HIGH-weight input for any score ≥ 2.
- AI-maturity false negative. A "quiet but sophisticated" company (no public signal but real AI work) will score 0. Note this in
ai_maturity.language_notes so the email drafter asks rather than asserts absence.
- Gap over-claiming. Do not emit a gap unless ≥ 3 of the cohort show the practice publicly. Sector-universal practices only.
When to hand off to a human
If any of these trigger, set requires_human_review: true and stop:
- Prospect appears in layoffs.fyi within last 30 days with ≥25% headcount cut (sensitive; tone risk).
- Leadership change is a founder departure or a publicly announced restructure (brand risk).
- Competitor gap brief would name a direct Tenacious client (from
delivery_bench_summary.md → do not use client names in a gap brief under any circumstance).
Bench-to-brief match
Before closing the brief, cross-reference the prospect's implied need against bench_summary.md:
- If hiring signals point to "Python data engineers" and the bench has zero available, emit
bench_match: {matched: false, gap: "python_data"}. The email drafter will then avoid specific capacity commitments.
- Never hallucinate bench capacity. The bench summary is ground truth.
