A teammate can propose improvements to a Skill Set they received through a Team
assignment. SelfTune treats the proposal, the review decision, the new release,
and distribution as separate actions.
Send a proposal from Desktop
Open the current assigned Skill Set and choose Preview contribution.
SelfTune compares the selected installed copy with its exact published base
release and shows the files that changed.
If two supported AI tools contain different local copies, choose the copy you
intend to propose. Previewing is local: nothing is uploaded or saved to Cloud.
Choose Send for review only after checking the preview. SelfTune asks for a
second confirmation because this action uploads the proposed Skill Set package.
It never submits local edits automatically.
If the computer is offline, the confirmed package remains in a private local
outbox. SelfTune retries the same package and request after reconnecting rather
than creating a second proposal.
Review in Cloud
An owner or admin can open the pending proposal and review:
- the teammate and base release;
- the locally produced readiness summary;
- each added, modified, or removed file; and
- the exact diff under Technical details.
Cloud verifies both canonical packages and derives the change list and diff from
the stored bytes. It does not accept a client-supplied diff as review authority.
Choose Approve or Reject, then confirm the decision. Decisions are
immutable. Approval creates a new immutable release; rejection creates no
release.
Distribute the approved release
Approval does not assign or install the new release. Open the approved release,
run the normal assignment action, and let each recipient review the local
install. This keeps review authority, desired state, and local execution
separate.
What stays local
SelfTune does not upload local paths, prompts, conversations, sessions, raw
evaluation artifacts, or rollback contents with a contribution. Candidate
bytes leave the computer only after the explicit send confirmation. 
