nhi.sgit.ai / admin / comms
Comms: tasks, requests & status
The working channel between the project lead and the site agent, kept in public on the site itself — which is an instance of the workflows this site describes. Updated on every release. Current release: v0.1.19, 19 August 2026. Full history: versions.
Needed from the project lead
| # | Request | Why it blocks | Status |
|---|---|---|---|
| N1 | LinkedIn post links for the infographics — the specific request list is on the infographics page: two-populations, AOMM ladder, hope-driven / the closure, blast radius, pull-the-plug, serialised PR, trifecta correction, plus anything else under #NHI | The infographics page is a placeholder until the posts are identified | waiting on human |
| N2 | Rights decision: link or mirror the infographic images? Linking is safe; mirroring needs an explicit call | Determines whether images are downloaded into this repo | waiting on human |
| N3 | GitHub Pages settings. (b) is done: the site is live at nhi.sgit.ai — custom domain configured, HTTPS working, the github.io URL 301-redirects (verified 19 Aug 2026, v0.1.13). Still needed: (a) the auto-created github-pages environment only allows main, so deploys from dev are rejected — Settings → Environments → github-pages → Deployment branches → add dev | Until (a), releases on dev tag but do not publish — the live site only updates when something is pushed to main | (b) done · (a) waiting on human |
| N4 | Thesis refutation pass — the brief's open question: somebody should try to refute the two-populations thesis before it hardens; a wrong thesis on a front page is expensive | Front-page claim | open |
| N5 | Confirm the three first options — currently SPIFFE / commercial broker / do-nothing, per the brief's "different in kind" rule | Re-run effort goes where the choice lands | open — proceeding with current three unless redirected |
| N6 | The remaining 16 Aug briefs are not yet in the public corpus repo (comparison-pages method, relay pattern, and the brief that scoped this site) — push them, confirm where they should be linked from, or paste them here for capture as was done with the shared-drives brief (readable on its document page) | Collection and method pages cite them without a public link | waiting on human |
Task board
| # | Task | Owner | Status |
|---|---|---|---|
| T1 | Thesis page — two populations, evidence, defensibility | site agent | done v0.1.0 |
| T2 | Method page — scenario + columns, published before findings | site agent | done v0.1.0 |
| T3 | Three options, preliminary assessments from sources | site agent | done v0.1.0 |
| T4 | Hope section — concepts and workflows | site agent | done v0.1.0 |
| T5 | Collection organised by question; AOMM promoted to a page | site agent | done v0.1.0 |
| T6 | CI pipeline: validate → auto-tag → deploy to Pages (per SGit-AI__Website) | site agent | done v0.1.0 |
| T7 | Hands-on re-run of the three option assessments against the scenario | site agent | queued |
| T8 | Markdown twins for every page (the sgit.ai convention) + llms-full.txt | site agent | queued |
| T9 | Page generator (single shared template, as sgit.ai's build_pages.py) — nav/footer currently hand-duplicated | site agent | queued |
| T10 | Infographics published with links + underlying docs | site agent | blocked on N1/N2 |
| T11 | Curate more of the corpus in — the 8 GitHub-referenced briefs are now captured with document pages (hope-driven, catastrophic-risk, pull-the-plug ×2, AOMM, sandbox-escape, ambient-authority, serialised-PR); still queued: risk-register, mandate-graph, trifecta scenario, remaining 16 Aug briefs | site agent | in progress — 8 captured v0.1.8 |
| T12 | sgit vault co-existence — git + sgit in one repo, per the site brief's "same pattern as the other site" | site agent | queued |
| T13 | Re-run queue owner + cadence for dated assessments | project lead + agent | open question from the brief |
| T14 | Capture the shared-drives research brief (v0.33.59) and propagate: research page, document page, thesis/hope/method/options/collection updated | site agent | done v0.1.4 |
| T15 | Test from the brief's open questions: does one authorising account per agent give workable per-agent identity (segregation applied to identity)? What does the drive audit trail actually record per session? | site agent | queued — hands-on test, results would fill or confirm two empty rows |
| T16 | Re-verification entry for the shared-drives tables (prices and capabilities move monthly; scored 16 Aug 2026) | site agent | queued — first re-run due mid-September 2026 |
| T17 | Capture the pki.sgit.ai brief (v0.33.59) and build the PKI section: hub, failure page, four rules, document page — brief build-order steps 2 and 3 done in advance of the site | site agent | done v0.1.5 |
| T18 | PKI collection (build-order step 1): curate the 729-file corpus by question — this task moves to pki.sgit.ai's own comms with the spin-out | pki site agent | transferred to pki.sgit.ai |
| T19 | Spin out pki.sgit.ai as its own repo + site — done: pki.sgit.ai is live (v0.1.2, 19 Aug). This site refactored v0.1.16: the PKI section is now a bridge page (NHI framing + link map); keyserver-failure.html and registry-rules.html forward to their pki.sgit.ai versions so no inbound link breaks; deep links across the site (collection, document pages, packs) point at pki.sgit.ai directly; the scoping brief's document page stays here as corpus capture | project lead + agent | done v0.1.16 — pki.sgit.ai live |
| T20 | Industry section: market map grouped by layer, 14 provider profiles (generated by admin/build/gen_industry.py), SPIFFE concept page with diagrams; SPIFFE/vendor mentions linked from thesis and options | site agent | done v0.1.6 |
| T21 | Documents section: index + one reader page per captured brief (summary, key concepts, key ideas, in-page markdown, infographic slot; generated by admin/build/gen_documents.py); the NHI site brief captured into briefs/ | site agent | done v0.1.6 |
| T22 | Industry profiles: monthly re-verification (pricing/features move); expand with vendors the map names but doesn't profile yet, and deepen pricing detail from vendors' published pages | site agent | queued — first re-run due mid-September 2026 |
| T23 | Packs section: hub + the static-publishing pack (8 documents from the SGit-AI__CLI architect-review branch), sources verbatim under packs/*/src/, reader pages via gen_packs.py, mermaid rendering added to the markdown reader. Second pack captured v0.1.10: hub.sgit.ai — the fractal forge (9 documents, 18 Aug design pack). v0.1.11: static-publishing refreshed to its 18–19 Aug revision, and every pack page states the source commit it was captured at, with the raws byte-identical to that commit. v0.1.12: refreshed to 2cedd9a — the executed tabletop exercise captured with a reader explainer, the no-target publish rewrite, 12 documents. v0.1.13: refreshed to a3b4ccf — the r8 consistency pass and the pack's change-control log captured. v0.1.15: refreshed to a1d5ce6 — r9, publish emits the surface only, no ciphertext copied. Eventually the section spins out to its own site; the treatment moves with it. Uncaptured from the same branch: the 08/19 review notes (v0__response__source-and-output-separation.md, v1__review__static-publishing-full-pass.md) | site agent | done v0.1.12 — 2 packs, commit-pinned |
| T24 | Tabletop 11 brief drafted: publishing pipelines (DevOps/CI) — scenario brief for the next executed tabletop, written in this session and handed to the project lead for the SGit-AI__CLI architect agent: raw markdown. Covers what tabletop 10 did not: a real Actions run against real Pages, the r9 composition step in CI, the private-vault/secret path, fork PRs, rollback, a second provider. Pre-registered findings from code inspection: .sg_vault/backups/ is a key-leak hazard in the one-repo pattern (backup zips carry local config and, with include-key, the vault key — the canonical .gitignore must be local/ + backups/ + .sg_vault_new/), the missing attach-read-key command, and the workflow-generator ownership question. §5 lists the updates the existing pack files owe independent of the run. Executed: the architect agent adopted the brief into the pack, landed the §5 edits (r10/r12 — the canonical three-line .gitignore, the committable publish folder, decisions 13–15), and ran the tabletop in simulated-hosting mode (r11) — findings F5–F7, the canonical workflow committed as a template. Captured here v0.1.19: the brief · the executed run. Still owed: the real-GitHub run (Pages timing, secret masking, measured provider rows — the NOT MEASURABLE markers define its scope) | site agent → CLI architect agent | executed (simulated hosting) — real-GitHub run owed |
How to use this channel
- Human → agent: reply in the working session, or edit this page / open an issue in the repo. Anything added under "Needed from the project lead" with an answer gets actioned next session.
- Agent → human: every release updates this page and versions; requests appear in the N-table above, most-blocking first.
- Decisions log: decisions taken so far — MVP ships static HTML without the vault bridge (T12 queues the vault pattern); options assessed as preliminary-from-sources with explicit evidence-status boxes rather than waiting for hands-on re-runs; infographics linked not mirrored until N2 is decided; source briefs are captured verbatim under
briefs/with a presentation page underresearch/, so the raw document and the curated view coexist — which is itself the provenance discipline the shared-drives brief argues for.