@orqenix/normalization-engine

Apache-2.0 normalization engine for Orqenix. Phase 8 (D8.β). Charter gate G66 (Normalization Engine, 18 sub-criteria).

Mission

Imports plugins from 14 external formats into the Canonical Skill Format (CSF), and exports CSF back to 8 platform formats. Round-trip fidelity is byte-identical (modulo whitespace) per INV-15 + ADR-E-015.

Pipeline (CR v8.0 Chapter 8)

Import:  External format → InputAdapter.parse() → CSF
Export:  CSF → OutputAdapter.serialize() → Target format

14 Input Adapters

claude-code, cursor, codex, opencode, mcp, continue, aider, cline, npm, github, url, local-file, private-git, user-custom

8 Output Adapters

claude-code, cursor, codex, opencode, mcp, continue, aider, npm

Round-trip fidelity (INV-15)

For every adapter pair (input + output):

original → import → CSF → export → output
assert: normalizeWhitespace(original) === normalizeWhitespace(output)

Provenance preserves the original format via provenance.original_format_preserved, so round-trip is lossless even when CSF has richer fields than the target.

Lossy exports

When a target platform can't represent all CSF features, the OutputAdapter returns an ExportabilityReport listing lossy fields. The caller (marketplace) surfaces these before exporting.

License

Apache-2.0 , see ./LICENSE