Investigate the two superpowers repos: detached forks, local checkout points upstream, version drift

[twistedmelonman]

Do-later: work out why two superpowers repos exist, which one is authoritative, and whether the local checkout should be repointed.

Surfaced while triaging twistedmelonman/dotfiles#231. Nothing here is broken today — the plugin loads and works. The concern is that the pieces disagree about which repo is the real one, which makes findings hard to attribute and made #64 harder to write.

What is actually on disk and on GitHub

thing	points at	note
~/Developer/superpowers (local checkout)	obra/superpowers	origin is the upstream, over HTTPS. At v5.0.5, commit 8ea3981.
smartwatermelon/superpowers (GitHub)	—	fork=false, parent=none, Issues disabled. Created 2026-07-31T18:17Z.
smartwatermelon/superpowers-marketplace	—	fork=false, parent=none, Issues disabled. Lives as a submodule under claude-config.
marketplace marketplace.json	smartwatermelon/superpowers	version: 6.2.0
plugin cache	—	holds 4.0.3, 6.2.0, 6.3.0

The three things worth understanding

1. gh issue list in the local checkout lists someone else's tracker.
~/Developer/superpowers has origin = https://github.com/obra/superpowers.git, so gh issue list there returns obra/superpowers' 312 open issues — Jesse Vincent's, not ours. It looks like a healthy local backlog and is nothing of the sort. No default remote is set (gh repo set-default is unconfigured), so gh falls through to origin.

There is no local checkout of smartwatermelon/superpowers anywhere under ~/Developer — a find over every git remote there turns up exactly one superpowers checkout, and it is the upstream one.

2. The superpowers repos read as detached forks, not fresh repos.
Both report fork=false / parent=none, yet smartwatermelon/superpowers carries upstream ancestry: Jesse Vincent's commits 44c9b2d6 and 3dcbd5c4 (the v6.2.0 release) sit beneath the first local commit 18821ef9 ("chore: remove forced SessionStart hook injecting using-superpowers (#1)"). The repo was created 18:17Z and that first local commit lands 59 minutes later.

That shape is consistent with fork-then-detach. Worth confirming, because it determines whether upstream changes can still be pulled in cleanly and whether there is any intent to track upstream at all.

3. Version drift across the three sources.
The local checkout is 5.0.5, marketplace.json declares 6.2.0, and the plugin cache has 6.3.0 on disk. Three sources, three versions. Related: #64 carries a finding that the marketplace dropped a pinned commit ref, so the plugin installs from HEAD while version: still says 6.2.0 — that pin removal is probably why cache and manifest disagree.

Questions to answer

• Is smartwatermelon/superpowers a detached fork of obra/superpowers? Was detaching deliberate?
• Should ~/Developer/superpowers be repointed at smartwatermelon/superpowers (with obra as an upstream remote), or is it deliberately a read-only upstream reference? If the latter, gh repo set-default should be set so gh commands there do not silently address the upstream tracker.
• Is there meant to be a local checkout of smartwatermelon/superpowers at all? Right now it is only ever touched through the plugin cache.
• Why is Issues disabled on both? For a real fork that is the GitHub default; for a detached standalone repo it is a deliberate setting. If it is just inherited default, enabling it would let Recovered review findings for repos with Issues disabled (superpowers, superpowers-marketplace, Instapaper-MCP, headroom) #64's 10 findings move to where they belong.
• Should the manifest version, the cache, and the checkout be reconciled — or is the drift expected because the pin was intentionally dropped?

Why this is worth doing

Attribution. During the twistedmelonman/dotfiles#231 drain, ten findings had to be verified against gh api and the plugin cache instead of a local checkout, because the obvious local path is a different repo. Attributing by the local checkout would have been wrong in a way that looked right — the upstream tree still has the hooks/ directory the findings say was deleted.

Related

• Recovered review findings for repos with Issues disabled (superpowers, superpowers-marketplace, Instapaper-MCP, headroom) #64 — the 10 superpowers/marketplace findings, currently parked because Issues is disabled on both
• ~/.claude/pending-issues/ has accumulated 43 unfiled review findings with no drain path twistedmelonman/dotfiles#231 — the triage that surfaced this

[twistedmelonman]

Triage pass 2026-08-24: three of the five questions are now settled by evidence

Ran the checks this issue asks for. The factual questions resolve; the
intent questions still need you.

Settled: yes, both are detached forks

smartwatermelon/superpowers contains the complete upstream history back
to Jesse Vincent's initial commit — not just the two commits noted in the
issue body:

dd013f6c Jesse Vincent  Initial commit: Superpowers plugin v1.0.0
2d20c951 Jesse Vincent  Fix SessionStart hook: add hookEventName field
f8bd448b Jesse Vincent  Fix SessionStart hook: use command type with script
fba0ec21 Jesse Vincent  Fix hooks.json structure: wrap in top-level 'hooks' key
a3690801 Jesse Vincent  Add clear and compact to SessionStart matcher

Paired with fork=false, parent=none, a full upstream history is only
reachable by forking and then detaching (or by a push of a cloned history,
which amounts to the same thing). Both repos were created within 2 minutes of
each other — superpowers-marketplace at 18:15:56Z, superpowers at
18:17:42Z — consistent with one deliberate setup session, not an accident.

Settled: the local checkout is the upstream, and gh is misdirected

$ git -C ~/Developer/superpowers remote -v
origin  https://github.com/obra/superpowers.git (fetch/push)
$ git describe --tags
v5.0.5-4-g8ea3981

Confirmed: gh commands run there address obra's tracker. This is a live
foot-gun regardless of which way you answer the intent questions below —
whatever you decide about repointing, gh repo set-default should be set so
gh issue list in that directory stops silently returning someone else's
backlog.

Settled: Issues-disabled is deliberate, not inherited

has_issues=false on both. GitHub disables Issues by default on forks,
but these report fork=false, so the default does not apply — the setting was
made explicitly, or survived the detach. Either way, enabling it is a
one-click change that would let dev-env#64's 10 findings move to where they
belong.

Still needs your intent — I am not guessing at these

• Was detaching deliberate? The evidence says it happened; only you
  know if it was intended, and that decides whether upstream tracking is
  wanted at all.
• Repoint ~/Developer/superpowers, or leave it as a read-only
  upstream reference? Both are defensible. If it stays as-is, set
  gh repo set-default regardless.
• Reconcile the 5.0.5 / 6.2.0 / 6.3.0 drift, or is it expected?
  Depends on whether dropping the pinned commit ref was intentional
  (see dev-env#64).

Recommendation

Enabling Issues on smartwatermelon/superpowers and
smartwatermelon/superpowers-marketplace is the highest-value single action
here — it unblocks dev-env#64's 10 parked findings and costs nothing. That is
your click to make; I have not changed any repo settings.