caoer115/contracts
1
0
Fork 0
Code Issues Pull requests 4 Projects Releases Packages Wiki Activity

mod/v1 (draft): the mod channel and the turn-end mod's frames #16

Open
caoer115 wants to merge 18 commits from agent/83f24248-mod-v1 into main
pull from: agent/83f24248-mod-v1
merge into: caoer115:main
Conversation 0 Commits 18 Files changed 2 +216
caoer115 commented 2026-10-22 04:05:10 +00:00
Owner
Copy link

Draft of mod/v1.md (status: draft; no frozen.lock.md rows): the mod seam at ccc-statusd 0cba578c (branch agent/83f24248-turn-end-mod).

  • § attach — mod_attach (v, session_id, pid, subset within turn-end/forks/subagent), the daemon's checks in order and each refusal's text, the accept, the channel's life.
  • § bridge — cccd mod-bridge attach (attaches again after a daemon restart, ends when its seat is gone, stops on a refusal) and cccd mod-bridge send; how the turn-end mod runs one bridge per session id, a /clear or resume included.
  • § routing — deliver() sends fork_request, cancel_fork, subagent_request and memory_state to the channel ahead of the held connection, with the same afterRoutedDelivery bookkeeping; what happens with no channel.
  • § frames, § forks, § subagents — the four frames the mod sends back, turn_end kept on the held connection, the debounce, the fork and its turn cap (turn-end-cap), the subagent spawn and its report.
  • § evidence — the named tests at 0cba578c, and what has no test.

Every claim is cited file:line at 0cba578c; the fields of the frames themselves stay in client/v1, the one-shot rules in oneshot/v1.

Draft of `mod/v1.md` (status: draft; no frozen.lock.md rows): the mod seam at ccc-statusd 0cba578c (branch agent/83f24248-turn-end-mod). - § attach — `mod_attach` (`v`, `session_id`, `pid`, `subset` within turn-end/forks/subagent), the daemon's checks in order and each refusal's text, the accept, the channel's life. - § bridge — `cccd mod-bridge attach` (attaches again after a daemon restart, ends when its seat is gone, stops on a refusal) and `cccd mod-bridge send`; how the turn-end mod runs one bridge per session id, a `/clear` or resume included. - § routing — `deliver()` sends `fork_request`, `cancel_fork`, `subagent_request` and `memory_state` to the channel ahead of the held connection, with the same `afterRoutedDelivery` bookkeeping; what happens with no channel. - § frames, § forks, § subagents — the four frames the mod sends back, `turn_end` kept on the held connection, the debounce, the fork and its turn cap (turn-end-cap), the subagent spawn and its report. - § evidence — the named tests at 0cba578c, and what has no test. Every claim is cited file:line at 0cba578c; the fields of the frames themselves stay in client/v1, the one-shot rules in oneshot/v1.
mod/v1: draft — the mod channel, its routing and the turn-end mod's frames (ccc-statusd 0cba578c)
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
c426f52f37
mod/v1: re-pin at ccc-statusd 6da1afe3 (PR #983, main merged in; no cited file changed since 0cba578c) and Claude Code 2.1.307; the evidence cites the recorded proof session
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
f42e8ec107
mod/v1: review pass 1 — re-pin at ccc-statusd 5a08b77e
All checks were successful
ci/woodpecker/push/woodpecker Pipeline was successful
Details
64904c1931 
- kinds: only claude-patched seats load the mods (ucc-<profile>, ucc-auto,
  ucc-sdk exec the patched native with --plugin-dir per mod); a live one-shot
  registration (claude-stock, a native kind) gets no channel.
- attach: new check row, after the held-by-another rule: a live stock seat's
  session is refused; a stock register ends a channel that attached first;
  the accept is written before the channel is recorded (the channel's first
  line); an accept that cannot be written attaches nothing.
- frames: a one-shot turn_end is dropped for a live stock seat and when the
  held connection declares turn-end; the mod awaits its turn_end send inside
  turn.complete, 2 s at most; cwd is $.session.root().
- forks: no fork and no answer before a main turn completed in the process
  (or since /clear or resume); tool.check allow for the mod's own forks.
- bridge: restart pause 1 s doubling to 30 s, back to 1 s on an accept.
- evidence: the three new daemon tests and the register.test.ts tests as
  named; frames.test.ts has seven tests; the untested list updated; the
  recorded traffic stated as recorded at 6da1afe3.
- every file:line and the source line at 5a08b77e (internal/core/,
  internal/modbridge/admit.go).
Merge remote-tracking branch 'origin/main' into agent/80888ab7-mod-v1
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
992f56e441
mod/v1: contract review pass 1 of #16 — wording (M1, M2, M3, M4, M5 erratum, L1-L6) and citations re-pinned at ccc-statusd 13d7ba27
Some checks are pending
ci/woodpecker/push/woodpecker Pipeline is pending
Details
ci/woodpecker/pr/woodpecker Pipeline is pending
Details
b39ece9000 
M1: the held subagent_request replays at turn_stop and turn_end, a spent one is reported lost, its sender hears nothing more.
M2: a claude-patched seat that loads no mod has none of the three families; uhappy local mode loads the mods, happy remote mode and a native started otherwise do not.
M3: an interactive seat reloads a changed mod; installs copy the mods to a per-install folder a running seat keeps.
M4: a fork's ask becomes an allow; rule and hook denies stand, and a settings deny rule is the named difference from the injection.
M5: client/v1 erratum extended to § extensions.
L1-L6 as the review gives them; L3 names what changed since the 6da1afe3 traffic run. H1's sentence in § subagents waits for its measurement.
client/v1 erratum: also § messages and § client messages, the rows of the moved families (contract review pass 1 of #16, Not checked item 10)
All checks were successful
ci/woodpecker/push/woodpecker Pipeline was successful
Details
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
24920a952c
mod/v1: held subagent_requests follow a session rotation (contract review pass 1 of #16, Not checked 6); citations re-pinned at ccc-statusd 0f718cf4
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
a03f4b9f1f
mod/v1: citations at ccc-statusd 1436b34f — the before-first-turn fork_request case is its own test, so § evidence names it on its own row
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
215feb2308
mod/v1 + client/v1 erratum: contract review pass 2 of #16 (leader 69f2ce94's rulings)
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
6f92805da9 
- M1: the client/v1 erratum adds § messages' two paragraphs (the four
  families on the mod channel, with session_id and no internal_id, whether
  or not a held connection exists; the injection drops them) and § window
  (claude-patched registers the narrowed subset).
- M2: § kinds states the loading by mechanism — a seat loads the mods only
  when its launcher passes --plugin-dir — and names ucc's install.sh beside
  the get-ucc wrappers (cited at ucc PR #57's head 04d0390 until it merges;
  0e27095 passes none); the erratum names the mechanism, not a launcher list.
- M3: measured, no reload (§ evidence, shared folder, run b13eab66 m3
  run-3): another seat's types/ write and same-bytes rewrites of tests/ and
  of the loaded register.js reload nothing; a changed loaded file reloads
  every seat on the folder. § bridge says so; its old "even with the same
  bytes" is narrowed to what was measured (B3's cp -R also rewrote
  plugin.json and hooks.json; which write reloaded is unmeasured).
- L1, L2, L3: the allow's limits, the settings-deny difference for the
  subagent, the report's origin and idle start.
- L4: § routing states the hold's re-check (ccc-statusd L4 code fix).
- L5: the no-snapshot path and the engine's own copy replacement.
- L6: § attach's withdrawal sentences after the attach replay; --seat-pid
  for a non-ancestor; a second attach's refusal is tested.

New ccc-statusd lines (subagenthold.go:309-334, the L4 test) are cited at
the #983 head that carries the fix; the whole file re-pins at #983's merge.
mod/v1: ucc install.sh citation is a placeholder until the ucc release line's mods change has a sha (leader 69f2ce94: #983 stays off main until ucc releases carry mods)
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
a33ae6db9a
All checks were successful
ci/woodpecker/pr/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
This pull request can be merged automatically.
This branch is out-of-date with the base branch
You are not authorized to merge this pull request.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin agent/83f24248-mod-v1:agent/83f24248-mod-v1
git switch agent/83f24248-mod-v1

Merge

Merge the changes and update on Forgejo.

Warning: The "Autodetect manual merge" setting is not enabled for this repository, you will have to mark this pull request as manually merged afterwards.

git switch main
git merge --no-ff agent/83f24248-mod-v1
git switch agent/83f24248-mod-v1
git rebase main
git switch main
git merge --ff-only agent/83f24248-mod-v1
git switch agent/83f24248-mod-v1
git rebase main
git switch main
git merge --no-ff agent/83f24248-mod-v1
git switch main
git merge --squash agent/83f24248-mod-v1
git switch main
git merge --ff-only agent/83f24248-mod-v1
git switch main
git merge agent/83f24248-mod-v1
git push origin main
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
caoer115/contracts!16
No description provided.