PMFKit
For engineers and contributors. User-facing documentation lives at /docs.

Release Scope Definition

Purpose

Define what is in PMFKit v1 and what is explicitly out. Scope discipline is mandatory until v1 is production-ready.

In Scope (v1)

  • Single URL input: User (or API) provides one product URL.
  • Structured diagnostic: Evidence collection → Evidence Layer → CMO (Report 1) → optional BD (Report 2) → optional COO (Report 3). Optional CDO from same crawl.
  • One primary report: Report 1 (discovery verdict, component analyses, DRL, recommendations) is the canonical output. Report 2 and 3 are optional follow-ups.
  • Save and export: Report saving (analyses, snapshots), PDF/JSON export.
  • Auth and usage: Sign-in for running analysis; usage limits; API keys for programmatic access.
  • Golden baseline regression: Key fields (stage, recommendation, component analyses, DRL) validated against baselines.
  • 5 real startup diagnoses: Coherent output on 5 real URLs as release gate (CMO-defined URLs and expectations).

Out of Scope (v1) — Do Not Build

  • Continuous monitoring: No scheduled re-crawls or trend reports.
  • Multi-input ingestion: No multiple URLs, no batch uploads, no interview uploads.
  • Analytics integration: No GA, Mixpanel, or other analytics ingestion for diagnosis.
  • Auto-roadmap builder: No generated roadmap or backlog from the engine.
  • Enterprise dashboards: No multi-tenant or org-level dashboards.

Do not add for v1: Monitoring, uploads, analytics ingestion, dashboards, feature expansion, pricing complexity. v1 = public diagnostic authority only.

Enforce via code review and this doc. Future phases in post-v1-roadmap.md.

Production-Ready Criteria

v1 is ready when all of the following are true. No release without them:

  1. Evidence layer is structured (categories, schema, single input to agents).
  2. All agents follow the contract architecture (versioned I/O, no duplication).
  3. FinalDiagnosisV1 schema implemented and used as the canonical final output shape (lib/schemas/final-diagnosis-v1.ts, lib/pipeline/to-final-diagnosis-v1.ts).
  4. Confidence scoring (0–100) and evidence coverage summary implemented and exposed in final output.
  5. Public extract mode implemented (generatePublicSummary in lib/public-summary.ts).
  6. 3 canonical case studies completed and published under docs/public-diagnoses/.
  7. 5 external startup tests produce coherent output (CMO sign-off).
  8. No agent re-crawls: orchestration and agents consume Evidence Layer or prior reports only (see agent-contracts.md).
  9. Evidence → inference → judgment separation enforced (see reasoning-standards.md).
  10. Output format is consistent (one aggregated diagnosis shape).
  11. PMF Stage framework is finalized (doc + code alignment).
  12. Homepage messaging matches engine output (CMO alignment).
  13. Report saving works with the canonical flow.
  14. Golden baseline regression tests pass.

Until then, not “live.”

Documentation discipline

Every architectural or output-shape change must update at least:

No undocumented logic. If it’s not documented here, it doesn’t exist.

Strategic reminder

We are not building an AI tool. We are building the default diagnostic lens for early-stage founders. Clarity > complexity. Signal > verbosity. Confidence > hype. Discipline > features.

Related Docs