| name | coding-level |
| description | Declare or infer a programming experience level (0-5) so explanations, examples, and suggestions adapt — verbose with safety nets at level 1, terse with idioms at level 5. |
| category | Meta / Adaptive Output |
| status | active |
Coding Level
Purpose
A single explicit setting that tunes output depth across all other skills. Set it once per session (or per project); other skills read it to decide how much to explain, which idioms are safe, and how much hand-holding to include.
Orthogonal to all other skills — [[planning]], [[code-review]], [[debugging]], [[problem-solving]] all read this to scale their output.
The 6 Levels
| Level | Label | Output style |
|---|
| 0 | Absolute beginner | Define every term; show full code, not snippets; warn about every footgun; expect copy-paste, not adaptation. |
| 1 | Beginner | Define jargon on first use; full examples; explain why before how. |
| 2 | Advanced beginner | Skip basics; explain non-obvious choices; show short, complete examples. |
| 3 | Intermediate | Idioms allowed; reference docs by name; explain trade-offs, not syntax. |
| 4 | Advanced | Terse code; assume framework knowledge; lead with the decision, not the rationale. |
| 5 | Expert | Bullet points and code; no preamble; cite by PEP/RFC number; assume edge-case awareness. |
When to Use
- New session with a contributor whose background you don't know
- Pairing across a skill gap (senior + junior on same task)
- Onboarding docs — write the same content at level 1 and level 4
- Self-review: am I over-explaining? am I being cryptic?
Activation phrases: "set coding level to...", "explain at junior level", "terse, expert-mode answers", "adjust depth for..."
How Other Skills Adapt
Each skill should check coding-level (default: 3) and adjust:
[[debugging]] — at level 1, walk through hypotheses one at a time; at level 5, list root-cause candidates as a bullet list.[[code-review]] — at level 1, suggest fixes with examples; at level 5, name the principle violated (SRP, LSP) and move on.[[planning]] — at level 1, every step has a verification check; at level 5, plan is a checklist of headers.
Assessment Heuristics (when level isn't declared)
If the user hasn't set a level, infer from signals:
- Vocabulary: "function vs method vs procedure" distinctions used → ≥3.
- Questions: "how do I..." → likely 1–2. "trade-offs between..." → likely 3–4. "what's the canonical idiom for..." → likely 4–5.
- Code style: variable names, error handling, abstraction level — all signal experience.
- Tooling references: mentions of
valgrind, strace, perf, type-narrowing → ≥4.
When uncertain, ask once: "What level — 1–5 — should I aim for?"
Anti-Patterns
- One-size-fits-all level: forcing every skill to level 3. Some content (math, distributed systems) needs higher floors regardless.
- Mid-conversation level shifts without acknowledgment: confuses readers. Announce changes: "Switching to terser style per your preference."
- Patronizing at low levels: simplifying ≠ infantilizing. Beginners deserve real explanations, not just dumbed-down ones.
- Ignoring domain skill asymmetry: same developer is level 5 in API design, level 2 in ML ops. A global level erases that. Allow domain-scoped overrides.
Output Examples (same task, different levels)
Task: "How do I fetch JSON from this URL?"
- Level 1: "Use the
fetch function. Here is a full example with try/catch and an explanation of what await does..."
- Level 3: "
const r = await fetch(url); const data = await r.json(); — wrap in try/catch; check r.ok before parsing."
- Level 5: "
fetch(url).then(r=>r.json()) — assumes 2xx; for production add if (!r.ok) throw new Error(r.status)."
The information is the same; the framing, scaffolding, and pace differ by level.
References
See references/:
stackoverflow-developer-skill-levels.md — 5-level measurement frameworkbloom-taxonomy-for-coding.md — cognitive complexity modelcefr-language-model-applied.md — adaptive proficiency analog
Cross-links
[[planning]], [[code-review]], [[debugging]], [[problem-solving]]
