AI_CONTEXT:LANCE_PROFESSIONAL_LESSONS:v1
Purpose: compact cross-project lesson ledger for future agents; update via .ai/skills/professional-lessons-learned after completed tasks.

Rules:
- Store generalized architecture/engineering/operations heuristics only.
- Do not store project-specific features, names, credentials, URLs, account IDs, implementation diffs, or biographical facts.
- Prefer compressed rule form: `ROLE_TAG: Prefer X when Y; avoid Z unless W.`
- Merge duplicates into stronger existing rules; do not append near-identical lessons.

Seed lessons:
- ARCH: Durable context should be indexed, not duplicated; entry files should point agents to authoritative docs.
- ARCH: Preserve future option value with seams and contracts, not premature full implementation.
- ARCH: Prefer maintained libraries for standard solved problems; avoid owning custom implementations unless product-specific requirements justify the long-term cost.
- ENG: AI-facing context should be concise, imperative, and optimized for fast rehydration over human-friendly prose.
- ENG: Separate post-task artifacts by purpose: task evidence, generalized professional heuristics, and project-specific durable state should not be collapsed into one memory file.
- ENG: Use installed library primitives directly for standard behavior before creating wrappers; wrappers should reduce duplication, not recreate or obscure library capabilities.
- OPS: Treat local/deployed environment parity as a config/IaC problem, not a source-branching problem.