Honest Mirror

The assessment uses a five-level competence framework — Aware, Beginner, Competent, Proficient, Expert — because vague labels like "above average" or "pretty good" don't tell anyone what to do next. Each level has a concrete description of what the person can and can't do. Most people asking "how good am I?" are at Level 2-3 and think they're at Level 3-4. The gap between "I can follow a tutorial" and "I can solve a novel problem with the same skill" is enormous and usually invisible from the inside.

The skill is built around evidence. It asks for concrete artefacts — code you've shipped, writing you've published, a recent situation where the soft skill got tested — and assesses what's actually visible in the sample rather than how confidently you describe it. If you can only describe (no artefacts), the assessment runs anyway but flags low confidence and explains why: people consistently overrate skills they describe versus demonstrate. A single sample gets a sample-size caveat. Three samples across different contexts gets a sharper rating.

The output names three things specifically. Strengths — what to keep doing, with the evidence that proves it. Gaps — what's wrong, what it costs you, and a concrete fix path with a realistic timeline. Not "your code could be cleaner" but "you're not handling errors, your function names are unclear, and your modules are doing three jobs each — refactor your last project to single-responsibility classes over the next six weeks". The blind spot — the one thing you don't know you don't know, which is usually the highest-leverage thing in the whole assessment.

The skill pushes back if you push back. If your self-rating doesn't match the evidence, it doesn't retreat to flattery — that breaks the entire value proposition — and it doesn't double down aggressively. It grounds the disagreement in the specific evidence and offers a practical test: "Try this challenge at your claimed level. If you can do it comfortably, I'll revise the rating." The hardest assessment to deliver is "you're competent but unremarkable" — it stings more than "you're bad" because it's harder to deny. The skill delivers it anyway when it's true.