Name Echo's seat, and unpin the card wording #771
No reviewers
Labels
No labels
move-to-repo
coilyco-bridge-deploy
move-to-repo
coilyco-flight-deck-agent-compose
move-to-repo
coilyco-gaming-eco-app
move-to-repo
coilysiren-inbox
move-to-repo
unknown
🔒⚠️📦⚠️🔒 SANDBOXED 🔒⚠️📦⚠️🔒
autonomy
async-consult
autonomy
epic
autonomy
headless
autonomy
live-collab
c#
priority
P0
priority
P1
priority
P2
priority
P3
priority
P4
role/ai
role/creator
role/design
role/director
role/engineer
role/exec
role/human
role/ops
role/qa
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
coilyco-gaming/sirens-echo!771
Loading…
Reference in a new issue
No description provided.
Delete branch "eng/name-echos-seat"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Echo's seat becomes
Echo (it). Deep keepsGem (they).closes #770
Merge this before coilyco-flight-deck/agentic-os#1052
That is not a preference. 2.20.0 breaks this repository as it stands, and this PR is the fix.
The card's wording moved from
Agent-compose assigned the ...toAgent-compose assigned **you** the ..., andValidateSystemPromptanchored on the whole first clause. Every composing role fails against a 2.20.0 bundle:It fails the image build closed rather than shipping a broken agent, which is the right direction — but the base-image bump would leave this repo unbuildable until this lands. I found it by baking all eight roles locally against the real catalogue, not by reading.
The anchor now matches the stable half of the line. Upstream owns that wording and may revise it again; a full-sentence anchor turns every upstream copy edit into a downstream build break.
Why the seat is not in request.kdl
agent/compose/request.kdlis one template that bakes every roster role —stage-compose-sources.shrewrites only itsroleline. Anidentitynode there would rename all eight, includingcreator, which is Deep's seat. So the seat lives in the stage script keyed by role, and onlyopshas one, because only Echo brings its own identity.Verified by baking:
The record can now see a rename
agent/rendered/roles/<role>.bundle.txtexists so a diff answers "what does this role actually pull", and it recorded no seat — a rename would have landed invisibly in the one file meant to make it reviewable. It now carriesSeat: Echo (it).Also in range, and worth a look
A week of agent-compose releases could not ship (agent-compose#281), so this is the first time their effects reach here:
stratsbecameexecupstream. Its record is renamed accordingly.boundary-*skills, now delivered as ordinary skills rather than implied.opsgoes 4 → 7 skills, which is the visible half of that.Verification
All eight roles baked against the real catalogue with 2.20.0 and validated;
go test ./...,policy-check, the prompt snapshots, the role-record check, and the full pre-commit suite all pass.To get 2.20.0 locally I upgraded through the tap (
brew upgrade agent-compose), which is the sanctioned path and was possible because the release pipeline is fixed. I did not build the container image; that happens in CI.🤖 Generated with Claude Code
Blocked, and the ordering I gave when opening this was wrong. Do not merge yet.
ci / image-buildfails here and will keep failing. The image bakes withagentic-os:release, which still ships agent-compose 2.19.0, and this branch needs 2.20.0 twice over:scripts/stage-compose-sources.shinjects anidentitynode, which 2.19.0 rejects asunknown node.Seat:line,execin place ofstrats, andopsat 7 skills rather than 4. The build regenerates the records and fails on any difference.ci / testpasses, because the Go suite does not bake bundles. Only the bake fails.The corrected sequence
I said "merge #771 before agentic-os#1052." That was backwards. The dependency runs the other way: this PR needs the new base image, so it cannot go first.
agentic-os:releaserepublishes.Step 1 exists so step 2 does not leave
mainunbuildable in the gap. Without it there is a window where the old anchor meets the new wording.What happens to this branch
It keeps the seat rename, the per-role identity injection, the role-record
Seat:line, and thestrats→execrecord rename. The anchor change in it becomes a no-op once #774 lands, so this will need a rebase onmainbefore it can merge. I will do that after step 3 rather than now, since the records will need rebaking against the real 2.20.0 image anyway.#770 stays open as the tracker for the rename; #773 is the atomic issue #774 closes.