← ClaudeAtlas

public-equity-investinglisted

Local, edition-agnostic equity-research methodology playbook for AlphaCouncil analysts. Use as the research method for the evidence roles (market_data, earnings_deep_dive, forward_expectations, sell_side_revisions, earnings_call_transcript, quant_factor, valuation_long_short, news_industry_management, management_industry_voices, insider_sec). On Codex this replaces or supplements the curated remote @public-equity-investing workflow; on Claude Code it IS the workflow (driven by WebSearch + WebFetch). No financial-data API or vendor key required.
Zhao73/alphacouncil-agent · ★ 3 · AI & Automation · score 71
Install: claude install-skill Zhao73/alphacouncil-agent
# Public Equity Investing (bundled playbook) This skill gives every AlphaCouncil analyst the same deep-research method **without** depending on Codex's curated remote `@public-equity-investing` plugin. It is a runnable method with gates, not a style guide. Evidence is gathered live via the agent's own web search (`WebSearch` + `WebFetch`). > Educational/research only. Not investment advice. Report missing data; never fabricate figures. ## When to use Whenever an AlphaCouncil evidence analyst runs. Each analyst applies the **sourcing ladder** and its **role focus** below, then returns the standard JSON evidence packet (see AlphaCouncil's Agent Output Contract). Keep JSON field names in English; write prose in the user's language. ## Sourcing ladder (per claim, in order) 1. **Primary document first.** Prefer the issuer/regulator over aggregators: - filings → SEC EDGAR full-text (`sec.gov`), 10-K/10-Q/8-K, 424B, Form 4; non-US → the local regulator / exchange filing (HKEX, SEHK, EDINET, etc.). - earnings → the 8-K Ex-99.1 press release and the IR deck / transcript on the company IR domain. - price/quote → the exchange or a named quote page, with the exact as-of timestamp. 2. **Dated, scoped search.** Run a primary-locator query (restrict to `sec.gov` + the company IR/exchange domain), then a recency query (`<ticker> <topic> <year>`), then ONE mandatory disconfirming query (`<ticker> guidance cut` / `downgrade` / `accounting concern` / `short thesis`). 3