Open research agents and analyst workflows connecting sustainability evidence to financial materiality, security valuation, portfolio decisions, impact and stewardship.

Institutional-quality buy-side research · Tradable decision support · Traceable evidence · Auditable workflows · Human judgment

Research philosophy: AI should expand research coverage without obscuring sources, assumptions, uncertainty, legal boundaries or analyst accountability. “Institutional-quality” describes process design and review controls—not independent certification, guaranteed compliance or investment performance.

flowchart TD
    E[Primary evidence and mandate] --> M[Financial materiality]
    E --> T[Transition credibility]
    E --> N[Nature and biodiversity]
    E --> H[Human rights and social value chain]
    E --> C[Carbon accounting and attribution]
    E --> B[Bond and credit diligence]
    E --> S[Sovereign and social-impact finance]

    M --> V[Valuation and investment case]
    T --> V
    N --> V
    H --> V
    B --> V
    S --> V

    C --> P[Portfolio and mandate review]
    V --> P
    P --> ST[Stewardship and controversy]
    P --> I[Impact and blended finance]
    ST --> IC[Investment committee / human review]
    I --> IC

The repositories are designed around inspectable research-process controls:

• Evidence lineage: source IDs, dates, document locations and entity/instrument boundaries.
• Financial transmission: sustainability issues must connect to explicit cash-flow, credit, valuation or portfolio assumptions when an investment conclusion is made.
• Reproducible calculations: deterministic helpers expose inputs, units and assumptions rather than hiding arithmetic in prose.
• Missing-data discipline: unknown or unverified evidence remains unknown; material gaps can block approval.
• Revision control: upstream changes can invalidate downstream analysis and prior review.
• Independent challenge: counterarguments, thesis breakers and unresolved issues remain visible.
• Human accountability: agents support research decisions; they do not self-approve investments or execute trades.

The objective is to move from:

Sustainability disclosure
        ↓
Material issue
        ↓
Financial transmission
        ↓
Security / portfolio implication
        ↓
Valuation, relative value or risk
        ↓
Catalyst / monitoring trigger
        ↓
Human investment decision

A sustainability claim is not treated as a trade thesis by itself. Likewise, a credible label, target or impact narrative does not substitute for credit quality, valuation, liquidity or portfolio constraints.

Across the agent suite, repositories generally include:

README.md          → answer-first overview
SKILL.md           → agent/skill entry point
AGENTS.md          → research-stage contracts
WORKFLOW.md        → orchestration and handoffs
sf_agent/          → executable workflow and calculations
schemas/           → structured input/output contracts
skills/            → bounded analytical capabilities
prompts/           → specialist research instructions
references/        → methodology and source controls
examples/          → synthetic worked examples
tests/             → regression and control tests
.github/workflows/ → GitHub CI validation

Investment analysis: Sustainable Finance · Climate Investing · Fixed Income · Credit Analysis · Fundamental Equity · Sovereign Debt · Impact Investing

Sustainability research: Transition Finance · Carbon Accounting · Nature & Biodiversity · Human Rights · Stewardship · SFDR / Sustainable Mandates · Blended Finance

Research infrastructure: Python · Agent Skills · Structured Evidence · Deterministic Calculations · Workflow Validation · Human Review

My general investment-research work is available in:

• Institutional Growth Equity Skills
• Sector Equity Agent
• Earnings Analysis
• Claude Equity Research Skills

Sector-specific healthcare and life-sciences research is maintained separately at hh-health-AI.

These repositories are research and educational infrastructure. They are not investment advice, broker systems, autonomous trading agents, independent audit opinions or certifications of regulatory compliance. Most workflows require a human or external AI host to retrieve authorized evidence and perform substantive research. Inspect each repository’s methodology, validation scope and limitations before relying on its output.

Evidence → economics → valuation → portfolio decision → human judgment