Investment Idea Screener
Find long and short stock ideas through quantitative screens, thematic research, and pattern recognition.
The skill asks for direction, market cap, sector, style, geography, and theme before proceeding, and describes screening, research, and presentation rather than trading, writes, or destructive actions. The repository README also stages outputs for professional review. However, it lacks data-flow disclosure, external-data permission boundaries, sensitive financial-data handling, source attribution requirements, explicit confirmation, rollback, and short-risk controls, so points are deducted.
The workflow is internally coherent, with defined screening metrics and an output structure, while the repository provides general plugin-validation and secret-scanning CI. The skill itself does not specify available data connectors, execution mechanics, input validation, abnormal-input handling, or diagnosable failure messages, and its key paths are not covered by skill-specific tests, so the static score remains conservative.
The skill has a clear stock-screening, thematic-research, and investment-idea use case, with trigger phrases, requested parameters, and a warning that screens are not conclusions. It does not define non-fit ranges, regional regulatory differences, Chinese-language support, mainland-China network fit, data-availability assumptions, or user-boundary conditions, so points are deducted.
The document is readable and progressively organized, with screening criteria, an idea-presentation template, and limitation notes. Repository context supplies an Apache-2.0 license, verified official provenance, general CI, and a contribution/update path. The skill lacks its own version, changelog, dependency/install notes, FAQ, named maintenance responsibility, and troubleshooting guidance, so points are deducted.
The requested deliverables—5-10 ideas, documented methodology, comparison table, prioritization, risks, and next steps—cover the core idea-generation task. However, metric values, valuation data, catalysts, and evidence require additional data access and human verification; the files do not demonstrate directly usable completed outputs, and static calibration caps this at 7.
The screening rules, thematic workflow, and output fields are auditable, and the repository includes general plugin-validation and secret-scanning workflows. There are no skill-specific key-path tests, executed examples, third-party result corroboration, or source citations, so only limited static verifiability is supported.
- The skill generates investment ideas and long/short candidates but does not define advice boundaries, suitability limits, jurisdictional differences, or mandatory human confirmation; screening results should not be used directly for trading.
- The output template does not specify data sources, timestamps, or citation formats for valuation, market data, catalysts, and risks; these fields require independent verification.
- External research connectors, paid-data dependencies, failure handling, and mainland-China network reachability are unspecified; when data is unavailable, the agent should report the limitation rather than invent values.
What it does & when to use it
This skill systematizes the search for new stock investment ideas across direction, market capitalization, sector, style, geography, and theme. It defines search parameters, applies value, growth, quality, short, or special-situation screens, and then conducts thematic sweeps when relevant. The workflow maps value-chain beneficiaries and considers what may already be priced in. Its intended output is a shortlist of 5–10 candidates with methodology, comparative metrics, thesis points, risks, catalysts, and research priorities.
Asks for direction, market-cap range, sector, style, geography, and thematic parameters; applies value, growth, quality, short, or special-situation criteria; researches themes by defining a thesis, mapping direct and indirect beneficiaries, distinguishing pure-play from diversified exposure, and assessing what is already priced in; presents candidate companies with market cap, NTM EV/EBITDA, NTM P/E, revenue growth, EBITDA margin, and FCF yield, followed by thesis points, catalysts, risks, and suggested next steps.
- An equity researcher needs new value or growth candidates within a defined sector.
- An investment manager wants a stock shortlist linked to themes such as AI infrastructure, reshoring, or aging demographics.
- A research team needs both long and short candidates with explicit catalysts and failure risks.
- An analyst wants to compare 5–10 screened ideas before choosing names for modeling or deeper diligence.
Pros & cons
- Provides distinct frameworks for value, growth, quality, short, and special-situation screening.
- Combines quantitative criteria with thematic value-chain analysis and catalyst identification.
- Requires methodology, comparison metrics, risks, and next steps alongside each idea.
- The source does not identify market-data or financial-data providers or demonstrate automated execution of the screens.
- Screen results are candidates, not conclusions, and require further fundamental work.
- Standalone installation, test coverage, and real-world validation outside the documented Claude environments are not specified.
How to install
The skill is located at plugins/agent-plugins/market-researcher/skills/idea-generation/, and the source does not document standalone installation. The repository README supports claude plugin marketplace add anthropics/financial-services, followed by claude plugin install market-researcher@claude-for-financial-services; in Cowork, add https://github.com/anthropics/financial-services through Settings → Plugins → Add plugin and select the relevant agent.
How to use
After installing the Market Researcher agent, submit a request using a documented trigger such as “stock screen” or “find ideas,” for example: “Run a stock screen for US mid-cap growth companies.” Supply the direction, market cap, sector, style, geography, and theme where applicable. The source does not document the underlying data providers or an automated screening command for this skill.