ProjectIntakeTower - morning control surface
Owner Decision Board
Die alte Lern-Grafik zeigt Scores, aber keine Entscheidungskette. Diese Ansicht sagt,
was gestern belastbar entstanden ist, was Rauschen ist, welche Human Gates offen sind
und welcher einzelne naechste Job sinnvoll ist.
500
gefundene Projekt- und Artefakt-Cluster
66
Human-Decision-Eintraege, nicht blind delegieren
1
erlaubter naechster Arbeitsanker: ProjectIntakeTower
Morning answer
What happened
Ein Read-only Intake hat Projekte, Dirty-Repos, Duplikate, Archive-Kandidaten und Jobs erzeugt.
Trustworthy artifacts
Owner digest, triage files, verifiers, and done handoffs for Zentrale, ACT and KomandoZentral.
Noise
Cache-Repos, alte Harness-Fixtures, Score-Listen ohne Entscheidung, pending jobs ohne neuen Human-Kontext.
Next one job
pit-komandozentral-46af91fa1c-review is done; next is an owner decision on outcome scope.
Trusted handoffs
Zentrale score 70
Dirty slice is coherent and targeted tests already passed. Next step is the documented final integration gate, not more features.
validation next
dirty repo
ACT-agentcommandotower score 69
Useful guardrail ideas, but dirty bootstrap work and human-owned operator state block autonomous continuation.
needs owner
merge candidate
KomandoZentral score 42
Clean supporting routing engine. Human decision needed because docs say outcome writing is out of Beta while UI/tests implement it.
scope decision
clean repo
Do not touch yet
Sidecar family duplicate cluster
Several high-score lines are merge/token-burn candidates, but they touch safety, hook, daemon or coaching direction.
human gate
scope creep
AgentIDE needs human
Ranked as a human decision item. Treat it as a possible UI dependency, not an implementation target.
no build
Global profiles Hermes / hooks
Do not change global Hermes, Claude, Codex, OpenCode profiles or process automation from this intake flow.
environment risk
Dirty repo pressure
/home/smlflg/.hermes44 dirty
Highest dirty count and many agent artifacts. Inspect only with explicit owner goal.
UEBER_AGENT 37 dirty
High score and high plan count. Likely important, but not a first move without stabilization intent.
AgentTestUniSupport 22 dirty
Active and possibly near product value, but still behind a human-gated dirty-state review.
Lanes for action
Lane
Meaning
Safe next move
Stabilize dirty repo Useful project, but current worktree may hide unfinished user work.
Read-only status and diff review before edits.
Consolidation candidate Valuable idea, not a separate product line to keep expanding.
Extract decision notes, do not port code yet.
Read-only worker job Bounded investigation can produce a handoff under jobs/done.
One worker, one job, no subagents.
Archive/cache noise Repo cache, fixture, old benchmark clone or stale artifact.
Manifest only; no delete/archive action yet.
Current human decisions
KomandoZentral outcome scope now
Choose whether active outcome saving is accepted Beta scope or must return to a read-only placeholder until Phase C.
ACT bootstrap state before ACT work
Decide whether to continue, park, or discard uncommitted GH Token Furnace harness bootstrap changes.
Sidecar consolidation defer
Pick a target only after explicit owner review because the cluster is safety-layer adjacent.
Source links
This board is a human-readable index over existing artifacts. It does not replace verifiers or worker handoffs.