QuantumBFS

Use when the user wants a numerical result (ED / DMRG / QMC / VMC / any method) checked against an exact solution, or names a model from the solvable-models catalog in a verification context. Trigger phrases: "verify against the exact solution", "check my DMRG/ED/QMC result", "benchmark <method> on <model>", "is this energy right", "does this match the exact answer", "sanity-check this against Bethe ansatz / Onsager / free-fermion solution". Also fires when the user names any of the 63 catalog models below in a verification context — not only when they use the word "verify": - T1 quadratic/free-particle: tfim-chain (TFIM), xy-chain, kitaev-chain, ssh-chain, kitaev-honeycomb, hofstadter-harper, haldane-chern, anderson-1d, harmonic-chain - T2 2D classical/transfer matrix: ising-2d-onsager (Onsager / 2D Ising), ising-triangular, dimer-kasteleyn, six-vertex, eight-vertex, hard-hexagons - T3 Bethe ansatz/Yang–Baxter: xxz-chain, heisenberg-xxx, hubbard-1d-lieb-wu (Lieb–Wu / 1D Hubbard), lieb-liniger, yang

Use when an exact diagonalization track, ED reproduction, full spectrum, symmetry sector, scar, ETH, level statistics, quench dynamics, finite-temperature ED (FTLM/TPQ), spectral function, interior-spectrum / MBL eigenstates, or finite-cluster oracle needs method-level route and tool selection.

Use when the user names or describes a harness-tracked quantum lattice model. Match user prose to one of: - transverse-field-ising (TFIM): quantum-critical Ising chain / 2D Wilson-Fisher - heisenberg: SU(2) magnet, AFM or FM by sign of J - xxz-chain: spin-1/2 XXZ, Bethe-ansatz integrable, Δ tunes FM / Luttinger / Néel - j1-j2: frustrated Heisenberg, J2/J1≈0.5 spin-liquid candidate - shastry-sutherland: orthogonal-dimer AFM, exact dimer phase, magnetization plateaus - spin-1-xxz: Haldane phase, single-ion anisotropy - aklt: spin-1 bilinear-biquadratic, exact VBS ground state - kitaev-honeycomb: bond-dependent exchange, exactly solvable Z2 spin liquid, anyons - spin-ice-pyrochlore: 2-in-2-out ice rule, Coulomb phase, magnetic monopoles - mbl-disordered-heisenberg: random-field XXZ chain, ETH-to-MBL transition - rydberg-pxp: blockaded Rydberg chain, quantum many-body scars, |Z2⟩ revivals - dissipative-spin-lindblad: open spin lattice, Liouvillian spectrum, steady states - potts-clock: q-state, first-order / cont

Use when the user wants to reproduce a paper's figure or main result. Triggers include "reproduce paper X", "redo the figures of Y", "reproduce arXiv <id>", "put this paper through the harness as a calibration target", "walk me through reproducing this paper", "beginner reproduction", "I don't know what size to choose", "explain while reproducing", or right after `/download-ref` lands a new paper.

Use when choosing or running QuSpin as the Python fallback for exact diagonalization, constrained-basis ED, spin-chain ED examples, or QuSpin setup failures.

Use when choosing or running XDiag.jl for exact diagonalization, symmetry-resolved sectors, Lanczos/Krylov calculations, or XDiag setup failures.

Generate a challenge completion report from a finished run in tracks/. Checks submission cleanliness first, then walks through each block interactively.

Use to render a run's HTML report — a reproduction proposal before compute, results after, or any structured run summary. Triggers include "render report", "build the report", "publish results", "share results", or a skill like `/reproduce-paper` composing its report page.

Use when running the autoresearch loop for a project whose survey and validator gates have passed — executes batches of attempts (one git worktree + LOG.md each, scored only by the validator under a hard time limit), then reflects, reports, and re-plans; continues autonomously while authorized_rounds remain, otherwise stops for user review. Stage 4 of the autoresearch pipeline.

Use when setting the goal and building the validator for an autoresearch topic — defines the publishable bar with the user, splits dev instances from a sealed gitignored holdout, builds a Docker-canonical strict environment (recorded fallback allowed), implements the validate CLI with rich JSON errors, and passes the strictness self-test against negative controls. Owns the validator gate. Stage 3 of the autoresearch pipeline.

Use when building the evidence base for a chosen autoresearch topic — maps the insight areas needed to propose new ideas, downloads references to cover them via download-ref, distills each area into research/INSIGHTS.md for user selection, builds a structured domain database and pinned reference implementations, and catalogs all algorithms/software in research/CATALOG.md. Owns the survey gate. Stage 2 of the autoresearch pipeline.

Use when starting or resuming an autoresearch project — reads research/STATE.md to find the current stage of the pipeline (topics → db → validator → run), verifies the previous stage's gate artifacts exist on disk, and routes to the matching stage skill. Triggers on "start autoresearch", "resume autoresearch", "autoresearch status", "what stage is my research project at".

Use when choosing research topics within a domain that suit autoresearch and defining their score metrics — brainstorms candidates, scores each on machine-checkable success, cheap per-attempt evaluation, iteration headroom, and publishability, lets the user pick, then derives primary/guard metrics (with computation cost and gaming risks) for each chosen topic and writes everything to topics.md. Stage 1 of the autoresearch pipeline; invoked via the autoresearch dispatcher or directly with a domain.

Use when building a slide deck in Typst + Touying for a scientific talk, lecture, or project briefing — ships a browsable zoo of color themes, layout templates, and gadgets (callouts, theory boxes, CeTZ diagrams, pin annotations) that compose into projector-safe slides. Distinct from slide-writing (HTML, QudeLeap brand) and paper-writer (full manuscripts).

Use when adding one or many new references (arXiv ID or DOI) to a sci-brain knowledge base — `<project>/.knowledge/` by default, or `advisors/<slug>/.knowledge/` when invoked from an advisor flow. Fetches metadata via Semantic Scholar, downloads PDFs (with SciHub fallback), renders to markdown, regenerates `INDEX.md`, and appends to the KB's `references.bib`. Handles both single refs and bulk-from-bib batches.

Use when brainstorming research ideas — a research collaborator that understands your background, helps find interesting problems together, and shares relevant resources along the way

Use when downloading a video or lecture from koushare.com (蔻享学术) — or a similar site whose pages are walled by a Tencent EdgeOne bot/CAPTCHA challenge while the video streams from Tencent VOD. Triggers/symptoms: curl on the page returns only a ~1KB EO_Bot_Ssid/__tst_status JS challenge, yt-dlp says "Unsupported URL", the page demands a "Security Verification" checkbox, or the stream URL contains voddrm.token / SimpleAES / a *.myqcloud.com .m3u8.

Use when restoring, reading, searching, summarizing, or resuming Claude Code logs/history from a session hash, UUID prefix, project path, or handoff/hands-off request. Also use when mining ~/.claude/projects JSONL transcripts for handoff notes, summaries, next steps, or lost conversation context.

Extract numeric (x, y) data points from a plot/figure image — including a figure inside a PDF — into a CSV, the CLI alternative to WebPlotDigitizer/automeris. Use whenever the user wants to "get the numbers off this plot", digitize/trace a curve, read data points from a chart or graph image, pull a dispersion/gap/curve out of a paper figure to compare against their own data, or recover tabular values from a figure when the underlying data isn't available. Triggers on phrases like "digitize this plot", "extract data from this figure/chart/graph", "read the points off", "get the curve from this PDF figure", "WebPlotDigitizer but from the command line".

Use when the user wants a live, auto-refreshing figure of cluster-job state pinned to their screen. Triggers on phrases like "babysit my X plot", "watch X live", "live view of <script>", "keep refreshing <figure>", "babysit scripts/foo.py", "open babysit". Also triggers on stop phrases: "stop babysit", "stop watching", "close the live view". This skill runs a project-owned plotter on the cluster on a recurring schedule, fetches the rendered PNG back to the local machine, and opens it. The plot updates in place as new data arrives on the cluster. Do NOT trigger for one-off plot generation ("run scripts/foo.py once"), status-only queries ("is job 12345 running?"), job submission, or analysis of a finished run — those are normal work.

Use when the user asks to download an academic paper or get the PDF — given a URL, DOI, paper title, or citation. Uses the bundled helper script as the default path for DOI/URL/PDF downloads, with open-web and arXiv lookup as fallbacks for title-only requests. Triggers on phrases like "download paper", "get me the PDF", "fetch this paper", "get this DOI".

Use when the user asks to mine their own Claude Code session history for recurring behavioral feedback — corrections, format/voice edicts, error catches, domain-knowledge injections. The output is normally proposed CLAUDE.md additions (the patterns are discipline rules, not workflows). Triggers on phrases like "distill my feedback", "find recurring corrections in my sessions", "what should I add to CLAUDE.md?", "/distill-feedback".

Use when the user asks to review a research-paper draft (typically LaTeX) or a related text (reply-to-referee, thesis chapter, technical note). Walks a fixed checklist of common issues — notation, figures, citations, structure, equations, style, reproducibility — and inserts inline author-comment reminders at locations that fail each check. Triggers on phrases like "review my paper", "check my draft", "apply the checklist", "review as a PRB reviewer".

Verify and validate bibliography references in BibTeX files by comparing with CrossRef metadata. Use when user asks to check, verify, or validate references.