You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Problem — The four director role cards (hypothesizer, experimenter, analyzer, implementer) exist only as deployed staging-bundle artifacts: unversioned, untested, no upgrade path — and the doctor's reverse-direction check flags machines carrying them as stale by construction. Approach — Author the role cards into the shipped agents directory as versioned, digest-verified registry artifacts owned by their mode bundles, SEEDED from the live deployed artifacts — with a human-gated seed diff against the engine-neutral definitions before any parity test pins content. Scope — in: the four role cards in-repo + registry ownership + staging/upgrade parity with the cards + the seed gate · out: skill moves (slice 3), plugin changes, any new role.
Acceptance Criteria
The four role cards exist as versioned repo sources, owned by their mode bundles, staged and digest-verified with the director cards
The doctor's source set includes them — a machine carrying the old deployed-only artifacts reads current once UPGRADED (and, if the seed gate amends content, reads stale until then — the condition is the honest outcome, not a regression)
The parity test (shipped card ↔ engine-neutral vault definition, where they overlap) passes on revision-pinned fixtures that carry the vault revision they pinned
The seed gate is recorded: the implementing cast prepared the diff, AARON performed and signed it, and the seed machine's provenance hash is in the campaign record — before any parity test pinned content
A scheduled pin-behind-HEAD check rides the doctor's fleet cadence on the vault-visible machine and files a chore issue on drift
Testing Decisions
Extend the staging/upgrade fixtures (role cards join the bundle) and the doctor staleness probe (the stale-by-construction verdict resolves). The parity test is new surface but composes the existing mode-card parity machinery.
Key Decisions
The seed is gated because deployed artifacts are plausibly per-machine drifted: garbage-in must never be enshrined as the tested baseline. Only THIS slice's parity pins block on Aaron's signature — no downstream slice waits on it.
Constraints & Invariants
The shipped card is the opencode binding; the vault definition is the engine-neutral contract — repo wins for shipped bindings, vault wins for engine-neutral semantics, and the parity test keeps the overlap coherent.
Prior Art
Spec decision D3; the registry slice's bundle layout; the engine-neutral role definitions in the private vault (agents records); the existing mode-card parity test machinery.
Source
Spec spec-20260905-063000-modes-first-class · Campaign ledger session-20260905-modes-first-class · Blocked by the mode-registry slice (bundle shape) and the ADR-amendment slice (publishing authorization — this is the first slice shipping previously-internal content). Published with Aaron's standing approval to execute (autodev, 2026-09-05).
Notes
Contains a HUMAN GATE (the seed-gate signature) — runnable to the gate autonomously; the signature itself waits for Aaron.
Important
Problem — The four director role cards (hypothesizer, experimenter, analyzer, implementer) exist only as deployed staging-bundle artifacts: unversioned, untested, no upgrade path — and the doctor's reverse-direction check flags machines carrying them as stale by construction.
Approach — Author the role cards into the shipped agents directory as versioned, digest-verified registry artifacts owned by their mode bundles, SEEDED from the live deployed artifacts — with a human-gated seed diff against the engine-neutral definitions before any parity test pins content.
Scope — in: the four role cards in-repo + registry ownership + staging/upgrade parity with the cards + the seed gate · out: skill moves (slice 3), plugin changes, any new role.
Acceptance Criteria
Testing Decisions
Extend the staging/upgrade fixtures (role cards join the bundle) and the doctor staleness probe (the stale-by-construction verdict resolves). The parity test is new surface but composes the existing mode-card parity machinery.
Key Decisions
The seed is gated because deployed artifacts are plausibly per-machine drifted: garbage-in must never be enshrined as the tested baseline. Only THIS slice's parity pins block on Aaron's signature — no downstream slice waits on it.
Constraints & Invariants
The shipped card is the opencode binding; the vault definition is the engine-neutral contract — repo wins for shipped bindings, vault wins for engine-neutral semantics, and the parity test keeps the overlap coherent.
Prior Art
Spec decision D3; the registry slice's bundle layout; the engine-neutral role definitions in the private vault (agents records); the existing mode-card parity test machinery.
Source
Spec spec-20260905-063000-modes-first-class · Campaign ledger session-20260905-modes-first-class · Blocked by the mode-registry slice (bundle shape) and the ADR-amendment slice (publishing authorization — this is the first slice shipping previously-internal content). Published with Aaron's standing approval to execute (autodev, 2026-09-05).
Notes
Contains a HUMAN GATE (the seed-gate signature) — runnable to the gate autonomously; the signature itself waits for Aaron.