Reviewed on 2026-09-28. This prototype is independently implemented. It does not vendor source code, prompts, model weights, UI assets or brand assets from the projects below, and does not depend on their runtimes. The list credits design inspiration, not endorsement or a claim that their capabilities were reimplemented.
| Reference | Idea informing this prototype |
|---|---|
| Agent Skills specification | A small, portable SKILL.md entry with on-demand reference material. |
| Codex Skills | Local/project Skill discovery; the host remains the coding agent. |
| Claude Code Skills | Same workflow in a client-specific discovery directory. |
| EvoScientist / EvoSkills | Persistent research knowledge and installable research workflows. |
| Xcientist | Research progress organized around inspectable evidence and stages. |
| SENPAI | Explicit experiment identity, artifacts and result contracts. |
| Beads | Agent-accessible durable state inside existing developer workflows. |
| TypeSafe / Jev patterns | Small typed judgments and programmatic control flow. No Jev model is used here. |
| DeepSeek Harness | Durable event records and replaceable integration boundaries. |
| ClearML | Everyday inspection, comparisons and provenance-oriented experiment UX. |
Licenses are not inherited by reading architectural ideas. Any future copied or adapted source must be reviewed against its actual license at the chosen commit, with the required notices retained. Do not silently copy third-party SKILL.md content and rebrand it. This file is a design bibliography, not a license audit.
The research-tree canvas also references the publicly visible Simate RoboScientist Experiment Gallery (Generalist Embodied Policy, viewed 2026-10-01): sparse status points, horizontal lineage progression, research-direction lanes and an on-demand inspector. The LabWeft layout and interactions are independently implemented. No Simate code, research data, images, brand assets or other site resources are included.
LabWeft is the confirmed project name; Research Harness was the prototype name. No claim of first invention, exclusive functionality, universal model support, or parity with the systems above is made.
The shinka extra now optionally installs
SakanaAI/ShinkaEvolve at commit
9912af12d423504b8d580f4179fd15f5f88b8c50, licensed Apache-2.0 upstream.
LabWeft's adapter is independently authored against its public Python interfaces
and native result schema. No Shinka source, prompts or assets are copied into
this repository. The earlier no-runtime-dependency statement describes the core;
it does not apply when this optional extra is explicitly installed. Retain and
review upstream license/notice material if distributing its dependency artifacts.
The trace investigation toolkit borrows design ideas from
morluto/rea (Reverse Engineer Anything, MIT,
reviewed at commit 9630c8ff46533fe63ec0e06212d01545070d17f0): summary-first
tools, three-valued results where absence is only claimed for exhaustive
coverage, explicit authority/confidence on recovered facts, reconciliation of
declared versus observed facts, an evidence-obligation ledger with stable
diagnostic codes, and unpacking of bundled artifacts. No REA source code,
prompts or assets are copied; LabWeft's implementation is independently written
in Python for research projects rather than binaries.