chrome-extensions.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: v0.1.0, 9 September 2026. Full history: versions.
Needed from the project lead
| # | Request | Why it blocks | Status |
|---|---|---|---|
| N1 | GitHub Pages settings. Confirmed by the project lead: the repo defaults to dev and Pages is enabled. Still to verify after this release deploys: (a) the auto-created github-pages environment allows deploys from dev — some repos in this network needed Settings → Environments → github-pages → Deployment branches → add dev; (b) the custom domain chrome-extensions.sgit.ai resolves with HTTPS once DNS + CNAME propagate | Until (a), releases on dev would tag but not publish | pushing v0.1.0 now — verify after first deploy |
| N2 | Confirm target-anonymity blocklist terms. admin/build/validate.js currently bans one term as a placeholder. The brief pack's own draft named a candidate once and caught it — worth the project lead reviewing whether the list should grow before more content is written | The failure mode this check exists for is a well-meaning editor adding a concrete example | open |
| N3 | Which extensions exist beyond the three efforts in this pack? The founder mentioned several "with different levels of quality"; this pack found three and graded them, but not ones it does not know about (brief pack gap G7) | The arc and the rubric can only grade what has been read | waiting on human |
| N4 | Code-snippet licence: CC0 or MIT? Currently stated as "CC0 or MIT" throughout pending a decision (brief pack Q4) | Every code sample on the site should carry one explicit licence | open — proceeding with "CC0 or MIT" until decided |
| N5 | Should rubric.json also ship as a skill on skills.sgit.ai, with this site owning the content and that one the packaging? (brief pack Q5) | Determines whether an agent picks up the rubric automatically before writing an extension | open |
| N6 | Can the recorder's source vault be reached for a verification pass, so the rubric's grades become measured rather than reported? (brief pack G2, G5, Q3) | The rubric's own grades of the recorder are currently asserted from the guides, not computed from source | waiting on human |
Task board
| # | Task | Owner | Status |
|---|---|---|---|
| T1 | CI pipeline: validate → auto-tag → deploy to Pages (per SGit-AI__Website), plus a target-anonymity blocklist specific to this site | site agent | done v0.1.0 |
| T2 | Front page — the five sentences, the narrow-instrument framing, the one-shipped-four-specified ratio | site agent | done v0.1.0 |
| T3 | The four guides shipped with navigation and cross-links, light editing only (/method/, /architecture/, /payloads/, /vaults/ scoped per deconfliction) | site agent | done v0.1.0 |
| T4 | The six silent failures, promoted to their own page | site agent | done v0.1.0 |
| T5 | The checklist — day-saving details, cross-linked to the unbuilt findings | site agent | done v0.1.0 |
| T6 | The 34-check quality rubric, plus rubric.json, graded against the three efforts including this project's own | site agent | done v0.1.0 |
| T7 | The unbuilt: hub page plus six proposal pages (U1–U6), findings promoted into architecture/checklist cross-links | site agent | done v0.1.0 |
| T8 | The arc: March → August–September → September, correction shown in place | site agent | done v0.1.0 |
| T9 | Documents section: the four guides readable in-page, raw markdown twins under /sources/ | site agent | done v0.1.0 |
| T10 | Boundaries, ethics and participant disclosure page | site agent | done v0.1.0 |
| T11 | Markdown twins for every HTML page (the sgit.ai house convention) + llms-full.txt | site agent | queued — currently only the front page and the four guides have twins |
| T12 | Page generator (single shared template, as sgit.ai's build_pages.py) — nav/footer currently hand-duplicated across ~25 pages | site agent | queued |
| T13 | sgit vault co-existence — git + sgit in one repo | site agent | queued — MVP is plain git |
| T14 | Verify the recorder's rubric grades against its actual source, once the vault is reachable (N6) | site agent | blocked on N6 |
| T15 | Curate more of the 122-file corpus mentioning Chrome extensions or MV3 — this pack followed the strongest three threads (brief pack G7) | site agent | queued |
| T16 | U6's open design details (cross-origin host permission, bundled crypto, offline/conflict behaviour) — worth a brief of its own before any extension ships against it (brief pack Q10) | site agent + project lead | open question from the brief |
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: MVP ships static HTML without the vault bridge (T13 queues the pattern); the recorder's rubric grades are reported from the guides rather than computed from source until N6 is answered; the unbuilt designs are published as first-class content rather than an appendix, per the brief's own argument; target anonymity is enforced in CI rather than by review.