Assessment Methodology

Assessments only help if you know exactly how they were made. This page documents how our seven instruments are constructed, scored, and versioned — the same facts your agents read through the context layer.

Shared design

All six instruments follow one design:

• Format: exploratory self-report. You rate first-person statements ("I enjoy building, repairing, or configuring something tangible") on a five-point scale from not like me to very much like me.
• Items: original items written for ProfileClaw, informed by each framework's published constructs. They are deliberately short screenings, not full-length commercial instruments.
• Scoring (mean-v1): answers are averaged per dimension, producing scores from 1 to 5 with two decimals. No population percentiles, no normalization against a norm group — a 4.20 means "you rated these statements high", nothing more.
• Reverse items: each instrument mixes in a small number of reverse-scored statements to offset agreement bias.
• Versioning: every instrument carries a version (2026-08.1 today). If items change, the version changes — and results you already saved keep the version they were scored under.

The six instruments

| Instrument | Basis | Items | Dimensions | | --- | --- | ---: | ---: | | Interest patterns (RIASEC) | Holland's vocational interest types | 18 | 6 | | Work style signals | Big Five (OCEAN) trait model | 15 | 5 | | Work values | Common work-value inventories | 12 | 6 | | Skills & talents | Self-report skill clusters | 10 | 5 | | Color temperament | True Colors (Lowry, 1978) | 12 | 4 | | Behavioral style | DISC (Marston, 1928) | 12 | 4 | | Workplace judgment | Situational judgement (SJT) tendencies | 12 | 4 |

Each dimension has a one-line description of what it measures, and each landing page documents that instrument's framing in depth.

What we do not claim

This matters as much as what we do claim:

• Not clinical. Nothing here diagnoses, treats, or screens for any condition.
• Not a selection tool. The instruments are exploratory and self-report; they are not validated for hiring, promotion, or any consequential decision about a person.
• No percentiles. We do not compare you to a population sample, because we do not run one.
• Screening trade-offs. Short instruments trade precision for speed. They reliably indicate which dimensions lead for you; they do not replace full-length instruments a certified practitioner would administer.

How results reach your agents

Scores feed one structured profile, and agents only ever see the shaped output — never your raw answers:

• The context packet carries schemaVersion 1.0, per-assessment versions, and provenance fields, so a reader always knows which version produced a score.
• Reads are task-shaped (career, projects, learning, communication) and scope-limited (profile:read, assessments:read, evidence:read).
• The packet's policy block marks assessments as exploration-only and evidence as user-submitted-unverified, so downstream AI treats results with appropriate caution.

Changes to this page

When an instrument's items or scoring change, we bump its version and update this page. The changelog on each assessment landing page reflects the current version, and saved results always retain the version they were scored under.