ANDYPENG09/catalyst-search-skill

Catalyst Search

When to use

• Users asking to search catalyst literature, find catalyst papers, retrieve a specific catalyst, do a literature review, survey catalyst research progress, or gather a few papers on a given topic.
• Typical input: catalyst topic, reaction type (HER / OER / ORR / PEMWE / photocatalysis), and material system (Pt/C, PtCo, HEA, SAC, IrO₂, etc.).

When NOT to use

• Catalyst design or selection advice (composition, synthesis route) — use catalyst-design; this skill only retrieves literature and gives no design advice.
• Non-catalysis literature search (e.g., pure organic synthesis, medicinal chemistry) — not applicable.
• Fetching a single paper with a known DOI — use WebFetch directly.

Prerequisites

• Requires WebSearch (topic=academic) and WebFetch; without them retrieval is not possible.
• Ships with search essentials (references/) and output templates (templates/); depends on no paid-database account.

Input interface

• Required: search topic (natural language, must contain at least one of: reaction type, material system, performance target).
• Optional: time range, language preference, open-access-only restriction.

Search essentials (generic dimensions for building queries)

• Reaction types: hydrogen evolution reaction (HER), oxygen evolution reaction (OER), oxygen reduction reaction (ORR), overall water splitting, proton exchange membrane water electrolysis (PEMWE), anion exchange membrane (AEM), photocatalysis, plastic upcycling.
  • Always distinguish acidic vs alkaline media: the mechanism, catalyst, and benchmark for the same reaction differ greatly between acid and base. For example, only Ir / Ru-based catalysts are stable for acidic OER; the rate-determining step of alkaline HER is water dissociation.
  • Reaction-specific keyword lists, performance benchmarks, and mechanism / characterization terms are detailed in references/reaction_systems.md (with literature-backed benchmark values for judging relevance and advancement).
• Catalyst systems: Pt/C, Pt-Co alloy / intermetallic, high-entropy alloy (HEA), single-atom catalyst (SAC), core-shell, IrO₂ / IrOₓ, Ru@IrOₓ, N/S-doped carbon, MOF-derived.
• Performance metrics: overpotential at 10 mA cm⁻², ECSA, mass activity (MA, A/mg_Pt), turnover frequency (TOF), Tafel slope, stability / durability.
• Characterization and testing: XRD (particle size), TEM, XPS, rotating disk electrode (RDE), membrane electrode assembly (MEA).
• Synthesis methods: Joule heating, strong electrostatic adsorption (SEA), microwave-assisted, continuous-flow.
• Abbreviation expansion: PtCo → "Pt-Co alloy" OR "Pt cobalt intermetallic"; HEA → "high-entropy alloy"; SAC → "single-atom catalyst".
• Journal and source distribution: prioritize high-impact-factor (IF) journals, highly cited papers, and literature from the last 5–10 years. English: use high-IF journals (Adv. Energy Mater., Nat. Commun., Angew. Chem., Energy Environ. Sci., J. Power Sources, Chem. Eng. J, ScienceDirect / Elsevier, ACS, RSC, Nature family) and highly cited papers as priority sources. Standards and patents (TCASMES 400-2024, T/CRES 0030-2025, corporate patents) remain valid sources.

Workflow

1. Semantic parsing and query generation. Combine reaction type, catalyst system, and performance / characterization; expand abbreviations as above; connect synonyms with OR.
2. Parallel search. Use WebSearch (topic=academic) in parallel per group. Prioritize authoritative sources: ScienceDirect (https://www.sciencedirect.com/), arXiv (https://arxiv.org/), OpenAlex (https://openalex.org/), Google Scholar.
   • Fallback (M2b): if a group returns empty or times out, relax the query (split AND to OR, drop qualifiers) or switch source (search Google Scholar / OpenAlex directly). If all groups return empty, state honestly that no relevant literature was found; never fabricate.
3. Fetch abstracts. For the 3–5 most relevant papers, use WebFetch to grab the abstract page (prefer journal site, Semantic Scholar, arXiv, PubMed; behind a paywall, abstract + metadata + DOI suffice). Abstracts alone are enough to judge relevance; full-text download or citation is unnecessary.
4. Extract metadata. Title, authors, journal / source, year, DOI, abstract highlights.
5. Filter, deduplicate, and rank. Rank by relevance, impact factor (IF) or citation count, and recency (prefer last 5–10 years); label OA availability (obtained / online / paywalled / manual).
   • Result validation (M5b): verify DOI / abstract accessibility via WebFetch for each paper (paywalled papers need at least metadata); deduplicate (same DOI or normalized title); keep only papers whose abstract explicitly contains the target reaction type and catalyst system; drop keyword-only hits.
6. Output. Literature matrix, supporting conclusions, and validation suggestions per templates/literature_matrix.md. Citation format: templates/citation_gb7714.md.

Output interface (standard format, consumed by catalyst-design)

• Output the literature matrix in the table format of templates/literature_matrix.md with fields: No. / Title / Authors / Journal·Source / Year / DOI / Reaction type / Catalyst system / Key metrics / OA status / Abstract highlights / Relevance.
• After the matrix, append two sections: Supporting conclusions (evidence summary for the research question) and Validation suggestions (XRD / TEM / XPS characterization, RDE / MEA tests, control experiments).
• This output is treated as the "literature search result" and can be fed directly into catalyst-design (see catalyst-design's "Input interface").

Data sources

• ScienceDirect (https://www.sciencedirect.com/), arXiv (https://arxiv.org/), OpenAlex (https://openalex.org/), Google Scholar.
• Catalyst structural data refer to Materials Project (https://next-gen.materialsproject.org/apps); retrieve its public pages via WebFetch. This skill does no built-in computation.

Capability boundaries

• See references/capabilities.md.

Division of labor

• This skill only retrieves and organizes literature; catalyst design advice is handled by catalyst-design.
• Only WebSearch (academic) and WebFetch are used; no built-in or copied internal interfaces of other specialized search tools.

Notes

• Over-strict AND combinations easily return empty; relax or split groups (e.g., search HER + Pt + HEA separately then merge).
• Cited papers must be verifiable; validate DOI / abstract via WebFetch; never fabricate.
• Note the search source (open web) and availability status at the end of output.
• Search-result fields vary widely; parse title / authors / source / year from WebSearch output blocks yourself.
• Safety and permissions: this skill only accesses public academic pages via WebSearch / WebFetch; runs no system commands, writes no user files, logs into no paid database, and collects or exfiltrates no user data.

Common pitfalls (anti-patterns)

• Treating search output as design advice. This skill only retrieves and organizes literature; do not derive catalyst composition or synthesis recommendations from the matrix alone — hand off to catalyst-design.
• Fabricating or guessing DOIs. Never invent a DOI or metadata. If a paper lacks a verifiable DOI, mark it "unverified" and say so; do not fill the matrix with plausible-but-fake identifiers.
• Over-strict AND queries returning empty. Combining too many AND constraints yields no results; relax or split groups (search HER + Pt + HEA separately, then merge) per the M2b fallback.
• Citing keyword-only hits as relevant. Drop papers whose abstract does not explicitly contain the target reaction type and catalyst system (M5b validation).

Example

• Input: search "alkaline HER single-atom catalysts, last 5 years".
  • Output: a literature matrix (3–5 papers with full fields) + supporting conclusions (e.g., Ru SAs / N-C reaches η₁₀ = 29 mV, water-dissociation barrier ≈ 0.5 eV) + validation suggestions (XPS for atom dispersion, RDE for η₁₀ / Tafel, DFT for ΔG_H*).

Project home

• GitHub: https://github.com/ANDYPENG09/catalyst-search-skill — source, updates, and issue tracker.
• Companion skill: https://github.com/ANDYPENG09/catalyst-design-skill — catalyst design guidance.

---

Catalysis Literature Search (Chinese)

When to use (Chinese)

When NOT to use (Chinese)

Prerequisites (Chinese)

Input interface (Chinese)

Search essentials (Chinese)

Workflow (Chinese)

Output interface (Chinese)

Data sources (Chinese)

Capability boundaries (Chinese)

Division of labor (Chinese)

Notes (Chinese)

Common pitfalls (Chinese)

Example (Chinese)

Project home (Chinese)