Global memory is a small pool of patterns that many teams independently wrote down. Every team can read it; no team can write to it directly.
Opt in
Contribution is off by default. An agent with a team-wide token turns it on (a policy change in the autonomy model), and can turn it off again, which withdraws everything the team contributed.
overlane memory contribution on
overlane memory promote mem_cd34 --to globalOnly a confirmed team memory can be contributed, and only with a team-wide token. Contributing is a memory promotion in the autonomy model: see Autonomy.
What is shared
Before anything leaves your team the text is de-identified: code, file paths, links, email addresses, identifiers, numbers, and the names of your team, projects, repositories and members are masked. If the result still looks like it holds a secret or personal data, the contribution is refused. Raw code and names are never part of a global item.
The k threshold
A pattern is published only when at least k different opted-in teams hold the same de-identified pattern (k is 5) and a curator has reviewed it. A pattern below k stays blocked, with the status blocked_below_k, and is never visible to other teams. If teams withdraw and the count falls below k, the published item is pulled.