abrahamferga/scout
The discovery loop: inventory every product already built on Plenipo, map industry coverage, and rank unclaimed verticals by how well the platform's spine (chat-first, RBAC-before-the-model, approvals, audit, multi-tenancy, documents, RAG, connectors, channels) fits them — ending in a go/no-go opportunity brief. Enable when choosing what to build next; disable once a vertical is picked.
Rank unclaimed industries as candidates for a new AI-first product on Plenipo by testing two things in order: whether the work needs deep AI — multi-agent orchestration, long-horizon agentic workflows, document processing over messy artifacts, retrieval that has to cite, judgment under ambiguity — and only then whether the workflow matches the platform's spine (tenant isolation, RBAC-before-the-model, approval-gated writes, append-only audit, connectors, chat + WhatsApp). Rejects the verticals that only sound good, above all CRUD systems with a chatbot bolted on. USE FOR: choosing the next vertical, sanity-checking an industry someone proposed, producing a ranked shortlist with reasons. DO NOT USE FOR: inventorying existing repos (../scan-fleet) or competitive research on a chosen industry (/define:research-industry).
Clear a brand name before it is spent. Generate candidates that escape the naming grammar every vendor in the category has already mined out, sweep the cheap registries (slug, GitHub org, NuGet, npm, .com), then run the probes that actually catch a live commercial product — fetching the exact .com, an exact-phrase search, near-neighbour spellings, the trademark classes software sits in — and sort every hit into fatal, survivable or ignorable with its evidence. Ends with at most three survivors and their residual risk, for a human to choose. USE FOR: naming a product before the launch pause, checking a name someone proposed, deciding whether a collision is fatal, re-probing before the repo is created. DO NOT USE FOR: choosing the vertical (../find-industry) or arguing it is worth building (../opportunity-brief); naming modules, tools or permissions (/define:plan-product); and never for registering a domain, filing a mark, or creating the repo.
Deep-dive ONE shortlisted industry into a go/no-go brief: the named buyer and the pain in units, the incumbent landscape and why an AI-first entrant wins or doesn't, the killer approval-gated workflow traced end to end, the agent design that carries it and the single-prompt baseline it must beat, the document and connector surface, RBAC tiers mapped to real job titles, a v1 that is genuinely shippable, regulatory burden rated apart from build effort, and an honest KILL CRITERIA check. The last gate before committing weeks of build — it never decides. USE FOR: writing opportunities/<industry>.md, pressure-testing a favourite candidate, killing an idea cheaply while it is still only prose. DO NOT USE FOR: generating or ranking candidates (../find-industry), taking stock of repos already built (../scan-fleet), or the competitor and feature matrix that follows a go decision (/define:research-industry).
Inventory every system built on (or adjacent to) the Plenipo platform and produce FLEET.md — which repos are true products versus legacy pre-platform systems, which platform version each consumes, how many host seams each adopts, which industries are occupied, and where docs have drifted out of sync with the code. The discovery loop's first stage and the input to find-industry. USE FOR: taking stock before choosing what to build next, auditing platform-version drift across products, finding stale architecture docs. DO NOT USE FOR: ranking new industries (that is ../find-industry) or researching one industry's competitors (/define:research-industry).