Authoring Authorization

For integrators and users — anyone who needs to know what decides whether a write to
Temper is permitted. This is the per-resource authorization axis; the prior question (may you
use Temper at all?) is answered in The Trust Boundary.

Authoring is an explicit capability, never obscurity

A cognitive map has no owner column — authorship is wholly an explicit write grant. Team
membership grants read, not write. Knowing a map's id gets you nothing: without an explicit
write grant, every authoring call is denied.

This is deliberate. Gating writes on obscurity (a private id, a team membership) is illusory
when container-write already lets a co-author supersede a node by fold-then-recreate. The
honest model gates on the explicit grant and records the actual actor on every mutation.

The three predicates

Predicate	Question
`cogmap_authorable_by_profile`	May this profile author into this map? — explicit write grant only
`context_authorable_by_profile`	May this profile author into this context? — owner, reachable team member, or explicit grant
`can_modify_resource`	May this profile modify this existing resource? — owner, explicit grant, or container-write cascade

Ownership is not a grant from a team

kb_resource_homes.owner_profile_id is one profile, and it is the only access-bearing profile
key — originator_profile_id is recorded provenance and confers nothing. Both gates read the
owner directly: neither team membership nor a context share enters that arm. Ownership is not a
grant, so nothing revokes it in place.

The consequence is deliberate. Withdrawing a share withdraws only what the share conferred.
Removing someone from a team, or unsharing a context, does not move ownership — so anything they
still own keeps them as its owner, and therefore as its reader and writer. Ownership moves only by
an act that names a successor: temper resource reassign for one
resource, temper team reassign for everything one person owns in a
team's shared contexts.

So a departure surfaces what is still owned rather than sweeping it. temper team remove-member
reports the removed member's residual, computed by the same query the handoff moves, so the report
and the handoff cannot disagree. What actually ends a departing person's reach is the prior
question — admission — and it is answered before any predicate on this page runs: see
The Trust Boundary, and
Offboard a Departure for the sequence an operator runs.

The container-write cascade

Whoever may author a container may modify any node homed in it — unix directory semantics:
directory-write implies file-rwx. A cogmap co-author can create nodes and fold, retype,
reweight, or update nodes another principal originated, without a per-resource grant. This is
deliberate collaborative stewardship. Provenance is unaffected — the event ledger records the
actual actor on every mutation, so "co-author B folded A's node" reads truthfully.

Agent vs. human principals

Agents are gated identically to humans. An agent authors because it holds an explicit write
grant — the same mechanism a granted human would use — not by virtue of any team membership or
ambient authority. There is no agent-specific bypass. Every call resolves to one concrete
profile, and every gate evaluates that profile.

When a human drives an agent (an AI assistant over the human's authenticated session), the
session authenticates as the human's principal — there is no separate "assistant"
principal. The assistant's tool calls carry the human's rights, no more.

Authoring Authorization

Authoring is an explicit capability, never obscurity#

The three predicates#

Ownership is not a grant from a team#

The container-write cascade#

Agent vs. human principals#

Further reading#

Authoring is an explicit capability, never obscurity

The three predicates

Ownership is not a grant from a team

The container-write cascade

Agent vs. human principals

Further reading