| name | company-diligence |
| description | Build a source-linked company or vendor diligence packet using Diligence Stack reports, vendor notes, and optionally entitled equity research. Use for company briefs, competitive positioning, thesis updates, catalysts, risks, monitoring dashboards, or investment research. |
Company Diligence
Build a living thesis that connects technical or market mechanisms to business outcomes and observable proof.
Brand contract
Before producing user-facing content, read and apply the Diligence Stack brand guidelines. Use its color, typography, logo, citation, and link defaults unless the user explicitly requests different visual styling; its attribution and canonical-link rules always apply.
Research sequence
- Follow the
diligence-research catalog-first MCP workflow.
- Resolve the exact ticker and relevant categories from
list_catalog. Search the shared diligence-stack-reports corpus first.
- Search separately for the company, competitors, market architecture, risks, disconfirming evidence, and dated watch items.
- If
equity-research is authorized, use it as a private comparative source - not as the house view. Search across multiple publishers and recent dates; separate substantive analysis from disclosures and valuation boilerplate.
- Fetch the evidence behind material claims. Retrieve the full vendor packet or report sections when snippets omit mechanism, assumptions, or caveats.
- Reconcile Diligence Stack analysis, company claims, and external analyst views. Do not average disagreement away.
Analytical frame
For each thesis component show:
driver -> mechanism -> business implication -> strategic read-through -> observable signal -> confidence
Evaluate:
- market role and architecture control;
- product and workload fit;
- customer proof, concentration, and production conversion;
- revenue quality, pricing architecture, and unit economics;
- supply, delivery, and ecosystem dependencies;
- competitive response and substitution risk;
- optionality versus evidence already in the base case;
- catalysts, watch items, and explicit falsification conditions.
For enterprise AI companies, distinguish the model/runtime/interface layer, orchestration/workflow layer, and system-of-record/governance layer when relevant. Weight production conversion and sustained usage more heavily than pilots, benchmarks, or new-logo announcements.
Output
Use the company packet. Label every metric as public fact, primary research, licensed research, derived estimate, or analyst judgment. Preserve ranges and confidence; do not manufacture false precision.
