# Preliminary repository review

Reviewed: 2026-09-23. Scope: the local Markdown scaffold, including pre-existing
uncommitted content. External source claims were not independently verified.

The scaffold has a coherent purpose and useful separation between research,
vocabulary, patterns and implementations. Reasoning from forces and resulting
contexts is its central strength. It is not yet an evidence-backed pattern corpus.

## Findings

1. **Metadata is inconsistent.** `canon/PatternSchema.md` uses `conflicts_with`,
   while `canon/Relations.md` and `templates/Pattern.md` use `conflicts-with`.
   Forces have prose labels in the vocabulary but underscore IDs in the schema.
   Resolve serialization and identity before tools depend on them.
2. **Prerequisites and graph edges are ambiguous.** Schema `requires` values name
   conditions such as `authoritative_state`, whereas relation definitions permit
   patterns or their contexts. Examples use names instead of `FEP-*` IDs.
   Define target types and reference resolution.
3. **Evidence is prospective.** The source catalog lists families, not individually
   identifiable references. The Carniato case is a stub without a source URL,
   date/version or claim references. README's “research-backed” description should
   be treated as an ambition until supporting evidence is recorded.
4. **Maturity and scales need definitions.** Schema `status` and `evidence.maturity`
   both use `hypothesis`; the catalog uses `seed`. Catalog scales include `L3/L6`
   and `cross-scale`, whereas the template uses one scale. Define valid states
   and representations before automated validation.
5. **Generative usefulness is untested.** Pattern families contain only READMEs;
   six architecture records and four specialized sequences are stubs. There are
   no technology mappings. Test a connected pattern set on contrasting scenarios
   before claiming coverage or expanding the catalog.

## Follow-up

`workplans/FEP-WP-0002-seed-language-validation.md` tracks these findings.
The preliminary intent preserves the mission and adds current maturity,
bounded priorities, success criteria and open decisions.

There is no application runtime, dependency manifest or automated test suite.
Validation for this change covers documentation, placeholders, classification
and State Hub registration/projection consistency.