score_breakdown.skills
Skills
Analyst
Named tools, credentials, and required-stack overlap.
Hive Intelligence System · internals
How Hive actually reviews an application
Six specialists do not each get the same prompt. Hive pages the structured profile into fact and narrative slices, fans them out in parallel, then the Navigator composes a score, a fix list, or a letter — and skips anyone whose view of the profile has not changed.
run_hiveparallel
- 01ingest profile, resume, job, mode
- 02page → fact lane | narrative lane
- 03skip specialists already current
- 04run the rest in parallel
- 05Navigator → score | actions | letter
- 06borderline score, no Empath → +1
- 07reuse until the profile or job changes
run_hive()
A fan-out graph, not a single prompt
Hive pages context, runs only the specialists that are missing or stale, synthesizes, optionally spends one follow-up call, then reuses what is still current. Select a stage to inspect the contract.
ingest
Ingest
The structured profile is the source of truth for gaps. Raw resume text is context only — specialists are instructed not to invent missing skills from it. The Navigator still sees the full profile.
Contract
- structured profile
- resume text
- job description
- career stage
- mode: score | resume | cover letter
context paging
Every agent sees a different slice
Hive does not dump the full profile into six identical prompts. Fact roles get skills, history, and credentials. Narrative roles get story, projects, and supporting documents. Inspect a specialist to see which profile fields they read and the honesty rule they follow.
Fact lane
skills · history · credentials
Narrative lane
story · projects · docs
specialist · analyst
The Analyst
Named tools, licenses, credentials, quantified results, ATS keyword density. Preferred skills never become high-priority gaps.
- Lane
- fact
- Model
- HiveAI Model
- Docs
- Excluded — facts only
profile fields
- full_name
- skills
- work_experience
- education
- certifications
- languages
Honesty · Gaps come only from the structured profile. Bachelor's = BS/BA/Bachelor; years are computed from employment date ranges — not phrase matching.
analysis_type
Three graphs, one specialist cache
Score, resume actions, and cover letters share the same specialist pool. What changes is the Navigator prompt, the model, whether tokens stream, and whether the final artifact is cached.
- Roster
- All six specialists, with a possible reduced set for score-only runs
- Navigator
- HiveAI Model. Paid plans use higher capacity so a free result is never reused for a paid review.
- Streaming
- Buffered JSON
- Output cache
- Reused until the profile or job changes
If the first pass is borderline and Empath was not on the roster, Hive runs Empath once and re-synthesizes. Gaps on required skills dominate the number; preferred-only misses are a modest penalty.
Navigator returns
- composite score: 0–100
- breakdown: skills, experience, culture, positioning, gaps, differentiation
- classification: exactly one primary label
- a short fit summary
- how the number was composed
Hive Score
How the Navigator composes the number
Hive Score is a 0–100 composite with a six-key breakdown and exactly one classification. Required-skill gaps dominate. Preferred skills, culture, and a clever angle cannot buy back a missing license or core stack.
score_breakdown.experience
Experience
Analyst + Strategist
Tenure from date ranges, scope, and role trajectory.
score_breakdown.culture
Culture
Empath
Narrative, values, and work-style fit — not a substitute for missing tech.
score_breakdown.positioning
Positioning
Strategist
Seniority band, market language, and competitive seat.
score_breakdown.gaps
Gaps
Skeptic
Higher is worse. Counts substantive required misses, not soft-skill boilerplate.
score_breakdown.differentiation
Differentiation
Innovator + Optimist
Honest distinctive angle. Cannot offset critical required gaps.
Required-gap caps
Soft-skill boilerplate is not a gap when work history shows the competency. Caps apply to named tools, licenses, degrees when education is empty, and other must-haves.
- One critical required miss≈ 60–65
- Two such gaps≈ 50–55
- Three or more25–45
- Hard gates (license, clearance, empty education)< 40
One primary classification
The Navigator picks the headline story — it does not combine labels. If several apply, it chooses what most limits fit or hiring risk.
Confident match
confident_match
Skills, level, and trajectory align. No major substantive gaps.
Strong stretch
strong_stretch
Plausible, but must grow on skills, domain, or scope — not overqualification.
Ambitious reach
ambitious_reach
Core requirements are missing or the experience shortfall is large.
Position mismatch
position_mismatch
Craft may fit; the seat, title, or seniority band does not.
Industry / domain pivot
industry_domain_pivot
Same craft, different world. Ramp and context are the risk.
Comp band risk
compensation_band_risk
Level or scope implies pay expectations above the posted band.
Culture / work-style risk
culture_workstyle_risk
Hard skills are adequate; Empath/Skeptic flag fit as the headline.
Credential gap
credential_compliance_gap
A specific license, degree, clearance, or cert is the gate.
Early career · high potential
early_career_high_potential
Thin on years, not empty on ability. Distinct from ambitious reach.
reuse
Spend calls only when the slice changed
Hive is not a six-call tax on every click. Specialists are reused across Score and Cover Letter Studio, re-invoked only when their view of the profile moves, and bounded to a single follow-up when a score sits on the decision edge.
Only re-run what changed
Each specialist sees a different slice of the profile. A skills-only edit re-runs Analyst, Skeptic, and Strategist and leaves Empath and Innovator untouched.
Share across modes
A score run and a later cover-letter run accumulate specialists instead of starting from scratch. Required analyses are reused when they are already current.
Paid and free stay separate
Paid Hive Score uses a higher-capacity version of HiveAI Model. A free result is never reused for a paid review, and vice versa.
Cover letters stay live
The letter itself is always newly written. Specialist findings behind it are still reused when that specialist’s view of the profile has not changed.
Same cache across surfaces
Score a role once. Reuse the swarm.
Hive Score, resume actions, and Cover Letter Studio share specialist outputs. A skills edit only re-runs the fact lane. The letter is always newly written; the six analyses behind it are not.