Add Valuein connector for SEC EDGAR fundamentals + smart-money + forensic-audit

[rainer85ah]

Summary

Adds Valuein as a partner-built plugin under plugins/partner-built/valuein/. Valuein hosts a free + low-cost MCP server at https://mcp.valuein.biz/mcp providing SEC EDGAR standardized fundamentals, ratios, valuation metrics, smart-money intelligence (insider transactions + 13F + 13D/G), and forensic-audit scores for the US public-company universe (~105M facts, 17,000+ active + delisted tickers).

The plugin packages 3 high-level commands and 3 supporting skills that demonstrate the connector's value out of the box. Sample-tier access is unauthenticated, so reviewers can evaluate the connector against S&P 500 names without provisioning a token first.

Why this PR

The existing partner connectors (lseg, sp-global) require institutional data subscriptions ($12K-$25K/seat minimum). The agent-plugins (earnings-reviewer, market-researcher, etc.) currently have no free or sub-$500 option for SEC fundamentals — agents fall back to scraping EDGAR or reading the filings text directly, which loses standardization across companies + multi-period cross-sectional analysis.

This connector fills that gap. The same shape as lseg/sp-global, the same SKILL.md conventions, the same lineage-grounded outputs — just on a different data layer aimed at the indie quant / fintwit / Pro-tier analyst persona that doesn't yet have FactSet / S&P CIQ access.

Structure

Mirrors plugins/partner-built/lseg/ and plugins/partner-built/spglobal/ directory-for-directory:

plugins/partner-built/valuein/
├── .claude-plugin/plugin.json     # metadata + mcpServers
├── .mcp.json                       # HTTP MCP endpoint
├── LICENSE                         # Apache-2.0 (same as repo)
├── README.md                       # commands + skills + integrations + install
├── CONNECTORS.md                   # 21 MCP tools by domain category
├── commands/
│   ├── research-equity.md          # equity research snapshot
│   ├── forensic-audit.md           # earnings-quality red-flag brief
│   └── screen-and-shortlist.md     # factor screen + forensic gate
└── skills/
    ├── equity-research/SKILL.md
    ├── forensic-audit/SKILL.md
    └── screen-and-shortlist/SKILL.md

Registered in .claude-plugin/marketplace.json alongside lseg and sp-global.

Commands

Command	What it does
/research-equity	Equity research snapshot — fundamentals trajectory, quality ratios, peer positioning, capital-allocation scorecard, valuation summary.
/forensic-audit	Earnings-quality brief — partial Beneish M-Score (SGI + TATA + LVGI), Sloan accruals, solvency snapshot, amendment history, capital-allocation cross-check.
/screen-and-shortlist	Cross-sectional factor screen → forensic gate → ranked shortlist with citations.

Each command has a matching SKILL.md providing domain knowledge, tool-chaining workflow, and an explicit "what not to do" section. Every numerical output cites its originating SEC accession via the response's lineage envelope (source_filing + source_url → clickable EDGAR link).

Connector design

• Auth: Bearer token in the standard MCP Authorization header. No secrets in checked-in files. Sample tier is unauthenticated guest access.
• Tier gating: Sample (free, S&P 500 / 5yr) → SP500 Free (signup, S&P 500 / full history) → Pro ($49/mo, full US universe, 30yr) → Institutional ($499/mo, smart-money + foreign issuers + redistribution license). The plugin's user-facing tools surface tier requirements clearly in README.md and CONNECTORS.md.
• Lineage: Every numerical response carries a structured lineage envelope. Skills are written to surface source_url in the final output so a reader can one-click verify any number against the originating 10-K / 10-Q.
• No internal references, no hardcoded secrets, no Anthropic-internal URLs.

Validation

$ python3 scripts/check.py
OK — 81 file(s) checked, 0 issues.

Secret-scan + internal-reference-scrub patterns from .github/workflows/secret-scan.yml run clean against plugins/partner-built/valuein/.

Live MCP preflight

$ curl -X POST https://mcp.valuein.biz/mcp \
    -H 'Content-Type: application/json' \
    -H 'Accept: application/json, text/event-stream' \
    -d '{"jsonrpc":"2.0","id":1,"method":"initialize",
          "params":{"protocolVersion":"2025-06-18",
                    "capabilities":{},
                    "clientInfo":{"name":"preflight","version":"1.0.0"}}}'

{
  "result": {
    "protocolVersion": "2025-06-18",
    "capabilities": {
      "tools":{"listChanged":true},
      "prompts":{"listChanged":true},
      "resources":{"listChanged":true},
      "logging":{},
      "completions":{}
    },
    "serverInfo": {"name":"valuein-sec-edgar","version":"2.0.0"}
  },
  "jsonrpc":"2.0",
  "id":1
}

License

Valuein's connector code is Apache-2.0, matching the repo's root license. The plugin includes its own LICENSE file with the Apache-2.0 text and Valuein copyright notice — same pattern as plugins/partner-built/spglobal/LICENSE.

Test plan for reviewer

1. Validation passes: python3 scripts/check.py returns OK — 81 file(s) checked, 0 issues.
2. No secrets: gitleaks + internal-reference scrub patterns return clean.
3. Structure matches partner-built convention: byte-for-byte mirror of lseg/spglobal file layout.
4. MCP endpoint live: https://mcp.valuein.biz/mcp responds to MCP initialize with valid protocol version + capabilities + serverInfo.
5. Sample tier works without auth: skills are usable on S&P 500 names without provisioning a Valuein token first (sample bucket access is guest).

Happy to address any review comments — naming, scope of the initial 3 skills, additional commands, or anything else that would make this a better fit for the repo.