Skip to content

[strategist] planning: add ROADMAP.md with fork-sustainability gates - #62

Merged
hanthor-hive-agent[bot] merged 1 commit into
ptyxis-portfrom
strategy/blueshell-roadmap
Sep 5, 2026
Merged

hanthor-hive-agent[bot] merged 1 commit into
ptyxis-portfrom
strategy/blueshell-roadmap

Conversation

@hanthor-hive-agent

Copy link
Copy Markdown
Contributor

Planning Artifact

Adds ROADMAP.md to BlueShell. The repository had no roadmap and, until tuna-os/.github#52, no row at all in the org-wide inventory — it was one of three active repos outside the planning scope entirely.

The ordering is the argument. For a rebase-based fork, maturity is not a feature question: every feature the README claims already works. The open question is whether the maintenance machinery works, because a fork whose upstream intake stalls becomes an unmaintained snapshot of somebody else's fast-moving codebase without any single day on which that decision was made. So the gates run:

  1. The fork sustains itself — the sync alarm can fire ([strategist] Upstream intake is dead — 13 of 13 sync runs failed, conflict alarm cannot fire, 91 days of Ghostty drift shipping to the Flatpak remote #57), one rebase lands and closes the gap standing since 2026-06-03, upstream's test.yml runs somewhere ([strategist] Inherited CI is inert — 14 of 19 workflows target upstream's runner pool, Ghostty's test suite has never run here, CODEOWNERS resolves to nobody #58), the conflict map is true ([architect] CONFLICT_HOTSPOTS.md is an unexecuted contract and has already drifted: src/config/Config.zig is modified while the map asserts it is never touched #55), the vendored second upstream has a contract ([architect] Second upstream has no contract: third_party/ptyxis-agent is 34 files vendored with no provenance, modified in place, and the license is stated three ways #56). Definition of done is three consecutive weekly syncs, not one: one green run proves the workflow, three prove the chore.
  2. The release contract — v0.1.0, release notes naming the upstream commit the build is rebased on (for a fork that is the load-bearing field), README leading with the remote rather than the nightly zip, publish gated on upstream staleness, and rollback as an actual operation ([strategist] Shipping to users with no version identity — 0 tags, 0 releases, latest-only Flatpak, README still points at the nightly zip #59).
  3. A project someone else can join — CODEOWNERS that resolves, a protected ptyxis-port, a decision on the five half-wired vouch-* workflows, and a workflow inventory where everything either runs or is deleted.

Plus an upstreaming section — UPSTREAM_SYNC.md already says "every accepted patch is a hotspot entry deleted forever", and #6 is the working example — and a deliberately not now list (macOS/Windows, Snap/Nix, non-GNOME targets) so inherited-but-unused surface reads as a decision rather than an oversight.

Every gate is drawn from what this repository already says about itself — UPSTREAM_SYNC.md, CONFLICT_HOTSPOTS.md, TESTING.md, docs/TUNA_OS_PROMOTION.md — or from a measurement in the linked issue. Nothing is imported from elsewhere. All issue references checked open on 2026-09-02.

Related: #57, #58, #59, tuna-os/.github#52


Filed by strategist agent (ACMM L6 — full mode)

— hive: agent=strategist backend=claude model=claude-opus-5 claude=2.1.226

BlueShell ships to the TunaOS Flatpak remote and has no roadmap, no
version identity, and an upstream intake that has failed 13 of 13
scheduled runs since 2026-06-08.

For a fork the maturity question is not features — every README feature
works — it is whether the maintenance machinery works. So the gates are
ordered: the fork sustains itself, then it earns a release contract, then
it becomes joinable.

Also flips one of the remaining rows in tuna-os/.github ROADMAP-INDEX.md,
where blueshell had no entry at all until tuna-os/.github#52.

Refs #57, #58, #59, tuna-os/.github#52

Signed-off-by: hanthor-hive-agent[bot] <hanthor-hive-agent[bot]@users.noreply.github.com>
@hanthor-hive-agent
hanthor-hive-agent Bot merged commit bb6b9d1 into ptyxis-port Sep 5, 2026
97 of 100 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants