context-status
context-status is how you read the consigliere workspace.
digest-source fills it; this answers questions over it. Two weeks
after a meeting you ask “how’s payments? what’s blocked?” and get a
grounded answer assembled from the materialized state — not a re-read of
every transcript. It is read-only and strictly grounded.
One command
/octopus:context-status "<natural-language question>"# or a path: /octopus:context-status paymentsHow it answers
- Resolve + read scope. Reads
consigliere.workspace; every read stays inside it. It never writes. - Interpret + route. Maps your question to a context or project from
the tree — the same routing as
digest-source. Ambiguous? It asks, it doesn’t guess. - Read the materialized state. Reads the node’s
state.md(the six fixed sections). It follows a fan-out pointer into a project’s detail, or readsjournal.mdhistory, only when the question needs it — nothing more. - Answer, grounded, with the consigliere lens. Leads with exactly
what you asked — status, blocker, owner, risk — each claim carrying its
(src: …)provenance. It surfaces a relevant political risk or an approved heuristic (“owner tends to delay → consider a FUP”) as a grounded note, never as invented fact.
Not recorded means not recorded
If the answer isn’t in the recorded state, it says so plainly and points
you at /octopus:digest-source — it never fills the gap from outside the
workspace, and it never guesses (it reuses audit-grounding).
Where it sits
context-status is the consult member of the consigliere bundle,
reading the state digest-source writes and applying the consigliere
role. It feeds nothing back — the heuristics loop is playbook-review.