Principle Encode Lessons In Structure

Encode Lessons in Structure

Encode recurring fixes in mechanisms (tools, code, metadata, automation) instead of textual instructions. Every error, human correction, and unexpected outcome is a learning signal. Capture it, route it, and close the loop.

Why: Textual instructions are easy to miss. They require the reader to notice, remember, and comply. Structural mechanisms (lint rules, metadata flags, runtime checks, automation scripts) enforce the rule without cooperation.

Pattern: When you catch yourself writing the same instruction a second time:

1. Ask: can this be a lint rule, a metadata flag, a runtime check, or a script?
2. If yes, encode it. Delete the instruction
3. If no (genuinely requires judgment), make the instruction more prominent and add an example of the failure mode

Pick the strongest rung. When more than one mechanism would work, choose the strongest the situation allows (an unrepresentable state that cannot compile, then a lint or banned API that fails CI, then a canonical helper, then a runtime check), because agents copy whatever the surrounding code already does and a weaker guard becomes the next template.

Corollary: Don't paper over symptoms. If the fix is structural, ONLY use the structural fix. The instruction IS the symptom.

Feedback loop:

• Capture every correction. When the human intervenes or tests fail, decide if it's a one-off or a pattern.
• Route to the right layer. One-off -> brain note. Recurring fix -> skill or lint rule. Systemic issue -> principle.
• Close the loop. Don't just record. Apply now or create a concrete todo.

Anti-patterns:

• Acknowledging without recording ("I'll keep that in mind" does not persist)
• Recording without routing (a brain note about a lint rule that should exist is wasted unless the lint rule gets implemented)
• Fixing without generalizing (fixing one instance while leaving the recurring pattern intact)

Battle-tested: The same stale-doc correction was independently rediscovered twice in one day, in two unrelated sessions, before the real fix landed three weeks later — and even then, the fix was another text edit, not a mechanism tying the doc's claim to the code behind it. A textual correction that recurs is exactly the signal this skill names; catching yourself making it a second time (even in a different session) is the trigger, not just noticing it inside one conversation.

Where a mechanical check is realistic: only for claims narrow enough to carry a machine-checkable anchor — a doc line that names a specific behavior can require a comment or test filename it points to, so if that anchor's source changes, the doc line gets flagged as stale. The pattern that generalizes: engine/skills/draft-pr/scripts/validate-pr-body.mjs fails CI on a PR that claims visual verification but has no literal "Manually inspected:" line — structural because it checks for a marker string, not judged content. Most prose-level corrections don't reduce this cleanly; that half stays a judgment call.