A skill is Heron's method written down: the evaluation framework, the source-handling rules, your output template, and the quality gate every deliverable has to pass. The harness is what sits around the model, the tools it can call and the context it is given. Both run in your own stack, on the Heron MCP, and produce your deliverable rather than ours.
A skill is a packaged method: the instructions, Heron's evaluation framework, the source-handling rules, your output template, and a QA gate — everything an analyst would otherwise carry in their head. Here is one concrete job, a diligence memo on a named CEO, done both ways.
Same evidence, same rigor. The skill changes who holds the method and how many times it runs.
Every skill in the library is built on a Heron data source and produces one deliverable in your house format. The library is organized by the work it does, not by product.
Built and tested on 80,000+ Heron interviews. Applicable to any primary research interview, including transcripts your own team conducts.
One governs data access, so your AI resolves the right company and executive and reads the right interviews rather than guessing at them. One governs language, and gates every draft before it ships. Every report skill inherits both, which is why the outputs are consistent rather than eleven separate experiments.
Runs inside your firm's environment. Your prompts, templates, and outputs stay where your team already works.
Resolves companies and executives, retrieves interviews and ManagementTrack records, and returns them to your client with full source metadata.
Turns what the MCP returns — or any primary research interview you bring — into your deliverable, checked against Heron's quality standard before it lands.
“Eighty percent of my own Heron usage runs through the MCP and our skills library.”
Skills are opinionated. The opinions are Heron's twenty years of management diligence, written down. Clients receive the frameworks the skills are built on, so their own teams can extend them, adapt them, or build new capability on the same foundation.
How Heron organizes evidence on an executive around the questions the business turns on, rather than personality — and how confidence is earned from the sources rather than asserted.
How researcher-led management-diligence interviews are structured: source selection, question design, and the signals that separate a decision-useful call from a courteous one.
How outputs are graded before they ship: quote traceability, verdict support, gap disclosure — the same standard applied to human- and AI-produced work.
Available to Heron clients.