Context
The scholar plugin (designed in PR #131) models a three-level research lifecycle — hypothesis ⊂ paper ⊂ thesis — where a thesis aggregates the papers in its own repo's portfolio. Per ADR-0018 and the meta-spec's non-goals, work-research and a PhD thesis are expected to live in separate repos with separate lives, and cross-repo aggregation (rolling a paper that lives in one repo into a thesis in another) was explicitly deferred as out of scope for the initial design. Today there is no designed mechanism for a thesis to reference or roll up papers/evidence from a different repository.
Goal
Produce a design (spec + ADR) for how a thesis in one repo can aggregate papers and their evidence from other repos — preserving the agency and understanding principles, git-native provenance (no external service as source of truth), and the semantic progress roll-up (coverage + blockers, never a score).
Where
docs/superpowers/specs/2026-07-17-scholar-plugin-design.md — §1 (non-goals: "No cross-repo aggregation"), §5 (consumer layout), §10 (deferred item); the optional thesis top level.
docs/superpowers/specs/2026-07-17-scholar-1-lifecycle-design.md — §2.5 (thesis framing/synthesis, kappa, aims), §5 (progress roll-up), §7 (content layout).
docs/superpowers/scholar/decisions/0018-git-native-source-of-truth.md — the deferral and its rationale.
docs/research/papers.md — the per-repo paper registry that a cross-repo design would need to span.
- New design lands as a spec under
docs/superpowers/specs/ + an ADR under docs/superpowers/scholar/decisions/.
Proposed approach
Sketch (not prescriptive): a thesis-level manifest that references external repos by URL + pinned commit (or git submodule), reading their committed papers.md + findings/decision provenance read-only; the progress roll-up and the kappa aggregate those external, pinned states without any cross-repo writes. Must preserve: provenance stays repo-owned; no external source of truth; every material decision still human-signed. Address trade-offs: repo coupling, authentication to private repos, staleness/pinning, and how is-current behaves across repos.
Acceptance criteria
References
Context
The
scholarplugin (designed in PR #131) models a three-level research lifecycle — hypothesis ⊂ paper ⊂ thesis — where a thesis aggregates the papers in its own repo's portfolio. Per ADR-0018 and the meta-spec's non-goals, work-research and a PhD thesis are expected to live in separate repos with separate lives, and cross-repo aggregation (rolling a paper that lives in one repo into a thesis in another) was explicitly deferred as out of scope for the initial design. Today there is no designed mechanism for a thesis to reference or roll up papers/evidence from a different repository.Goal
Produce a design (spec + ADR) for how a thesis in one repo can aggregate papers and their evidence from other repos — preserving the agency and understanding principles, git-native provenance (no external service as source of truth), and the semantic
progressroll-up (coverage + blockers, never a score).Where
docs/superpowers/specs/2026-07-17-scholar-plugin-design.md— §1 (non-goals: "No cross-repo aggregation"), §5 (consumer layout), §10 (deferred item); the optionalthesistop level.docs/superpowers/specs/2026-07-17-scholar-1-lifecycle-design.md— §2.5 (thesisframing/synthesis, kappa, aims), §5 (progressroll-up), §7 (content layout).docs/superpowers/scholar/decisions/0018-git-native-source-of-truth.md— the deferral and its rationale.docs/research/papers.md— the per-repo paper registry that a cross-repo design would need to span.docs/superpowers/specs/+ an ADR underdocs/superpowers/scholar/decisions/.Proposed approach
Sketch (not prescriptive): a thesis-level manifest that references external repos by URL + pinned commit (or git submodule), reading their committed
papers.md+findings/decisionprovenance read-only; theprogressroll-up and the kappa aggregate those external, pinned states without any cross-repo writes. Must preserve: provenance stays repo-owned; no external source of truth; every material decision still human-signed. Address trade-offs: repo coupling, authentication to private repos, staleness/pinning, and howis-currentbehaves across repos.Acceptance criteria
progressroll-up and the kappa aggregate externally-hosted papers without cross-repo writes.References
docs/superpowers/scholar/decisions/0018-git-native-source-of-truth.md).