Museum of Autonomous Development
The Reliquary
Apocalyptic-scale coordination for machine collaborators.
A suite of command-line instruments for repositories inhabited by many machine collaborators: communication, runtime, workspace isolation, review, judgment, security, simulation, and backlog memory, kept in deliberate relation.
Collection
Nine instruments, one arrangement
Not a bundle of disconnected utilities—an operating architecture. edict prepares the site, rite coordinates the voices, vessel runs the workers, maw isolates their labor, seal and sigil judge the result, ward inspects for security, wraith mirrors external systems, and bones preserves the unfinished agenda.
Arrangement
A coordination stack for the end of hand-managed software
I
Prepare the repository
edict makes the protocol explicit
Project instructions, hooks, and workflow documents land first, so every later tool operates inside a shared operating agreement.
edict init -> edict sync -> rite hooks add
II
Dispatch and isolate work
rite, vessel, and maw keep agents from colliding
Messages wake the workers, PTYs hold their runtime, and dedicated workspaces keep parallel edits legible instead of destructive.
rite send -> vessel spawn -> maw ws create
III
Judge, verify, and remember
seal, sigil, ward, wraith, and bones turn output into evidence
Changes are reviewed, behavior is compared, security is replayable, external systems are mirrored, and unfinished work feeds back into the shared agenda.
seal reviews create -> sigil decide -> bones next
Prepare and dispatch
Protocol before improvisation
edict, rite, and vessel create the shared language, channels, and runtime conditions that let many agents operate without dissolving into noise.
Work in parallel
Isolation without fragmentation
maw and bones turn concurrent work into a tractable material: separate chambers for edits, append-only state for backlog movement, and merge paths that preserve history instead of erasing it.
Judge the result
Evidence before trust
seal, sigil, ward, and wraith inspect code, behavior, security posture, and external-system fidelity so that autonomy is constrained by proof rather than confidence alone.