[zheng-weihao]

What

Clarifies what belongs in the memo card, across skills/orbit/SKILL.md, skills/CONSTRAINTS.md, docs/spec-knowledge.md, and evals:

• Card scope unchanged — two questions (roles + MVP/VIP entry points). Deep code structure is now phrased as "not the card's business" with no prescribed destination (previously "belongs in a code-doc", which nudged agents to create documentation unprompted — colliding with the general no-docs discipline).
• Repo facts only — facts about the user, collaboration, current effort, or external systems never go into the pool-shared card. The boundary keys on what a fact describes, not who learned it or when, so entry points discovered mid-task still count as card material.
• CONSTRAINTS.md's card example fixed to the scaffold's actual two-section structure (## When to add (roles) / ## How to use).

Evals

• fix: cruise hint skip condition — enumeration-recall test, not content-in-context #32 wording synced to the new phrasing.
• New docs: auto-approve tier contract — judgment-based restatement, framework-neutral done/new #34: a pitfall goes to neither memo nor jot.
• New fix: read paths never fetch — info and session-start hook go purely local #35: an entry point discovered mid-task is card material, not per-agent memory.

No code changes; evals.json validates.