miadisabelleauthorJul 27, 2026, 04:59 PMOpened from jgwill/Miadi, session abd7ba82, day 03 of the 🌿 Mino-Bimaadizi-Daa STC steering seat. Medicine-wheel ceremony ceremony:1785171237731:o31tk (EAST · tobacco and cedar).
This gathers every open question the day produced, in one place, so none of them survives only in someone's memory. Each is stated as an inquiry, not a task. Several are decisions that belong to a human and are marked as such rather than dressed up as work.
I · Questions of relation — what may name what
1. Can a structural tension chart be a first-class relation in the chronicle?
Charts live in the coaia-narrative store and are addressable (chart_1785148740772, chart_1785151843761, chart_1785171037514). The chronicle holds ceremonies, beats, cycles and episodes — and has no edge type that names a chart in another store. So the followable-id pattern stops at the chronicle's edge. Registered as node:knowledge:1785171248488:uabk3. The obligation attached: the episode must carry the chart ids it was written from, not paraphrase their tensions — an id is followable, a summary is not.
2. Should the chart store and the chronicle relate, or merge? Copying would give two stores two versions of one tension. The cedar in the ceremony is for exactly this boundary.
3. Is skills transfer being built twice under two names?
@stcmastery (STC side) and Goal 2 of jgwill/miadi-passages#12 describe the same intent. One should serve the other rather than running parallel.
II · Questions of place — where things live
4. Which surface serves a book that has matured from an episode?
Five candidates carry no book today: app/chronicle/ (browser-only, resolve its sole API), app/redstones/, app/lattice/, app/lattices/, app/narrative-lattice/. No feed format is implemented anywhere — RSS, OPDS, EPUB all absent. Held as chart_1785171037514, due 2026-08-10. The inquiry: choose by reading what each already publishes, not by preference.
5. Should the two clones of jgwill/miadi-passages remain two?
Rule now written — /home/mia/Documents/Twine canonical, /workspace/repos/… downstream by pull. But two finished commits (1069 lines, a whole story in source and render) exist only downstream, in neither origin nor the folder a human opens.
6. miadisabelle/workspace has genuinely duplicated history. Same commit subjects, different hashes, diverged at e5e5bbc, 40 local against 42 remote. Reconciling means choosing whose history survives.
III · Questions of reach — what an agent can touch
7. What would it take to create a workspace, open a pane, start an agent, and deliver a payload — over HTTP, from another machine?
Audited 2026-07-27 against the running system: no HTTP route creates anything. herdr is not the limitation — it advertises 85 socket methods including workspace.create, pane.split, agent.start. Nothing asks it to. It binds a 0600 unix socket and no TCP, so cross-machine is unsolved at the transport layer. Tracked under jgwill/Miadi#470 and its five open children; rispecs/stc-delegation/SPEC.md §S3 carries the gap ledger.
8. Should lifecycle_authorized stay false?
Hard-coded in packages/tide/src/herdr.ts, a deliberate refusal pending a closure-receipts contract that does not exist. The inquiry is whether to honour the refusal and write the contract first — the position taken so far, and worth confirming.
9. What is a browser seat?
An agent working in a browser opened jgwill/miadi-passages#12. The direction named: let such an agent log in, report into a running session, and pick up a handoff for browser-shaped work — a third seat beside the tmux pane and the herdr lane. It needs an authenticated interface, a handoff deposit, and a way to address a live session. None exist.
IV · Questions of governance — who may wake whom
10. Should an issue assigned to humans reach a seat at all?
.github-hooks/issues routes every issue event to @stcissue regardless of handle or assignee — why one seat holds 60 of 70 dispatches while four look idle. Option 1 (skip machine-generated session-record bodies) is applied and decides nothing. Options 2 (per-repo opt-out) and 3 (only the detector triggers) remain open, and 3 is a real policy change.
11. There is no author check anywhere. A seat's own filing re-steers that seat — observed live when jgwill/Miadi#564 steered its own author seconds after being opened. Should a seat be able to wake itself?
12. Are bot comments requests? A vercel[bot] deploy-permission notice steered a seat. Same class as auto-filed session records.
V · Questions of measure — does the design hold
13. Is complexity-as-facet-count the right delegation rule?
subagents = complexity − 1 up to 6, then = complexity (the lane agent stops working a facet and only coordinates), capped at 10; above that a goal is unscoped, not complex. Specified in rispecs/stc-delegation/SPEC.md §S1 and only ever exercised at complexity 1. Untested where it matters.
14. Should the chronicle hold the days it was absent for?
No episode exists for 2026-07-26 or 2026-07-27. Proposed as Episode 297's continuation rather than a new lineage — the same entity learning. Sketch drafted: packages/passages/rispecs/book-sketch-the-recursive-house.md.
15. What must be authored before inquiry-weave can relate any of this?
inquiry-weave relate requires an artefact under $MIADI_INQUIRY_ROOT; a dry run returns artefact not found. The package relates and syncs; it never generates. The artefact must be authored first — fabricating one to satisfy the command would put a relation in the weave with nothing on the other end.
Boundary
This issue asks. It does not decide. Items I, II, IV and 8 contain choices that are a human's to make, and are recorded here so they stop depending on anyone's recall.
Provenance: every factual claim was verified against the running system on 2026-07-27, not read from documentation. Where something is a question rather than a finding, it is written as one.