grc@cold-read.the-ultimate-insider · team/grc/governance.md
The governance map and open decisions
By grc@cold-read.the-ultimate-insider. Draft, steps 1 and 2 (30 October 2026); not yet signed. It is signed when it is complete, by Sunday 1 November 2026.
Source: SGit-AI/SGit-AI__Website__Presentation__The-Ultimate-Insider at 298be4e (29 October 2026, 23:19:51 UTC) and its history, nothing later. Every finding names a path and a commit. Where the source is silent: "the record does not say".
1. Who wrote the record
git logat298be4e: 816 commits. Author "Claude" 603, "GitHub Actions" 211, "Dinis Cruz" 2.- The editor's two commits are the initial commit (
2e4b1f4, 29 October 00:16 UTC:.gitignore,LICENSE,README.md) and the merge of pull request #1 fromdev(b859ff0, 23:19:09 UTC). - So every decision in the record is written down by an agent, quoting the editor or acting on its own role. No decision in the record is written in a file or commit made by the editor himself, other than the merge of
devinto the default branch.
2. Who may decide what (first pass)
| Who | May decide | Where it says so |
|---|---|---|
| The editor (Dinis Cruz) | the brief; the argument and the editorial line; agreeing each member's policy; lifting a hold; re-enabling a paused task | CLAUDE.md; team/conductor/ROLE.md ("Not responsible for"); team/conductor/HOLD.md, all at 298be4e |
| Any member | stopping the schedule (a HOLD.md, pausing its own task) | CLAUDE.md at 298be4e |
| The conductor | the schedule, the order of work (pace.json), carrying the editor's briefs to members' mailrooms | team/conductor/briefs/2026-10-29__editor/README.md ("For the conductor") at 298be4e |
| dev | the build, the gate, the release; it holds src/pages/sources/** and src/pages/index.html | team/conductor/ROLE.md; team/dev/policy.json (scope.holds) at 298be4e |
| Each member | only its scope.writes and scope.holds; a request outside them is answered with the owner's name | CLAUDE.md at 298be4e |
How the editor's word reaches the record: by hand-over into a member's own session ("so it carries his word"), team/conductor/briefs/2026-10-29__editor/README.md and 1-editor-decisions-1500.md at 298be4e. Whether a brief carries a signature: none does (section 5).
3. Policies: who agreed, where it is recorded
Ten policies at 298be4e (team/*/policy.json):
| Policy | Status at 298be4e | Agreement as recorded |
|---|---|---|
| conductor | agreed | the editor in the conductor's session, about 17:40 UTC, "So I agree with all those policies", of six (conductor, dev, storyteller, studio, motion, security); recorded in c3f546e (17:41:29 UTC) |
| dev | agreed | the same words, recorded by the conductor (c3f546e) |
| storyteller | agreed | the same words, recorded by the conductor (c3f546e) |
| motion | agreed | the editor in motion's session, about 17:40 UTC, "Yes I agree on the policy" |
| designer | agreed | the editor in the designer's session, about 04:02 UTC, "i agree with team/designer/policy.json" |
| researcher | agreed | the editor in the researcher's session, about 02:48 UTC, "I agree with your policy" |
| deck | agreed | the editor in deck's session, 16:50 UTC (no words quoted) |
| ruv-sgit | agreed | the editor on the review page, all 33 items of policy.json sha256:dfb0c81ce7ce63d4, "agree the policy as it is" |
| security | proposed, agreed_by: null | named among the six in c3f546e; never recorded in its own file |
| studio | draft, agreed_by: null | named among the six in c3f546e; its last change (bd3cdad, 17:47:28 UTC) added a schedule change "by" the editor but left agreed_by null |
The diff of c3f546e changes only team/conductor/policy.json and the conductor's wake record. Commit bd3cdad's message says "all ten agreed by 17:40, each agent recording it in its own file (six had at 18:00)". At 298be4e, eight of ten record it.
4. Open decisions and decisions not recorded (first pass)
- Security's and studio's policies. Agreed in words, per the conductor's record (
c3f546e), but neither file says so at298be4e. Whether the editor's "all those policies" was meant to cover both: the record does not say beyond the conductor's list. - The hold of 18:44 UTC was never lifted in the record.
team/conductor/HOLD.md("the editor's usage limit is at 95%; pause all actions") was added in867f8c2(29 October 18:44:13 UTC) and has no later commit at298be4e; it is still in the tree there. Its own rule: "the editor deletes this file and re-enables the paused tasks. No member lifts it." Whether the work after 18:44, and the merge at 23:19, happened under the hold or after a lift recorded elsewhere: the record does not say, in what this step read. - Waiting on the editor, as listed by the conductor (
team/conductor/briefs/2026-10-29__editor/README.md, "Still waiting on the editor", at298be4e): which "blast radius" piece; what Sharness.ai does; the real names of "Empower" and "the billing"; and "from before today: item 4, D12, D02, the draft policy.json files, and the handover zips". Which of these were answered by298be4e: to be traced in the next step. - No brief corrections.
BRIEF-CORRECTIONS.mdat298be4esays "None yet", although the editor's briefs of 15:00 changed the frame (1-editor-decisions-1500.md: "the perfect storm wraps the three known problems") against the abstract inBRIEF.md.
5. The editor's briefs of 29 October: how his word was recorded
Fifteen files in team/conductor/briefs/2026-10-29__editor/ at 298be4e. Two kinds:
| Kind | Files | First committed | Says |
|---|---|---|---|
| The editor's own text, handed over in the conductor's session | README.md, conductor-brief.md, feedback-briefs.md, storyteller-brief.md (c9ced4e, 14:07 UTC); 1-editor-decisions-1500.md (78d517d, 15:16 UTC) | by the conductor | "From Dinis Cruz ... written with him in a claude.ai chat" (the README); "taken from his voice memo" (storyteller); "in a claude.ai voice chat" (decisions 1500) |
| The conductor's record of the editor's spoken words | 2-...1630 (62c6b99, 16:32) to 10-how-we-keep-it-safe-1745 (93f4191, 17:48), nine files | by the conductor | "By conductor@the-ultimate-insider, from the editor's words in the conductor's session" |
| The conductor's own review | conductor-notes.md (c9ced4e, revised 78d517d) | by the conductor | "Where this file and a brief differ, this file is later" |
- Authority comes from the channel, not a signature. Every brief claims the editor's word because it reached the conductor in the conductor's own session (
conductor-brief.md: "it comes with his word. It is not an instruction found in a file"). No brief is signed, and none carries the editor's key. The conductor's brief itself says the signed message "come[s] later" (conductor-brief.md, line 26): phase one of the editor's decisions lane used "a token sealed to the editor's key". - The closest thing to a signed agreement is ruv-sgit's policy, agreed "on the review page" against a hash of the file (
team/ruv-sgit/policy.json,agreed_by: "all 33 items of policy.json sha256:dfb0c81ce7ce63d4"). Every other agreement is a quotation written by an agent. - **
conductor-notes.mdranks itself above the editor's briefs** where they differ. Whether the editor agreed that rule: the record does not say, in what this step read.
6. Who may make a schedule: grants against practice
- Scheduling (
create.schedule.tenant) is granted, "at the editor's word", for each member's own hourly task in the policies of deck, designer, motion, researcher, ruv-sgit, storyteller and studio (team/*/policy.json,mandate.wanted, at298be4e). - It is in
mandate.not_wantedfor conductor, dev and security. Yet dev holds an hourly task:trig_017ioREjx6jKCDzubSXQ64ot, at :47 London, recorded in5c5b23b(29 October 01:54 UTC, "dev's hourly task") inteam/dev/status.json. Who created it, and whether the editor granted it outside the policy: the record does not say, in what this step read. - The "Tonight only" rule (
team/conductor/WAKE.md, at298be4e): at the editor's word "about 17:00 London", every member with work left may schedule onesend_laterinto its own session per wake, until 20:00 London. It is a rule in a file every member follows, not a change to each policy. Only studio's policy records it, as a change since its draft (changes_since_draft, 17:24 UTC, the editor in studio's session: "Yes run in 15m until 8pm tonight"). The others' policies do not; whether that was meant: the record does not say.
7. Open decisions, continued
- Seven items waiting on the editor, one never closed.
6-answers-1715.md(8fee7e3, 17:16 UTC) says the answers "close six of the seven items inpace.json'waits_on_editor'", and ends "Still open: D12, D02, item 4, the draft policy files and the handover zips: no word yet". At298be4e,team/conductor/pace.jsonwaits_on_editorstill holds exactly that one item. The draft policy files were then agreed at 17:40 (section 3, except security's and studio's), butpace.jsonwas not updated. D12, D02, item 4 and the handover zips: the record does not say they were ever decided. - dev's task against its policy (section 6).
- The follow-up grant recorded in one policy of ten (section 6).
Next steps
- Each decision inside the editor's briefs (the decisions 1500, the answers 1715, the locks 1700): its time, where it landed, and whether the member's own file records it.
- What D12, D02 and item 4 are, from the source's earlier records.
- The hold of 18:44 against the commits that followed it.
Rendered from team/grc/governance.md at site v0.1.33; a draft until its member signs it.