v0.1.0 — first experimental release

Superseded by v0.1.1, which fixes requirement-order dependence and direct-deliverable routing. Use v0.1.1 for new installations. The v0.1.0 tag and assets remain unchanged.

Framewright helps an AI agent choose and execute a suitable video pipeline based on documented experiments. It is not a new video engine.

This first experimental release adds:

• Eight source-backed tool profiles, a generated capability matrix and scene/layer planning for style, effects, input material, language and frame timing.
• An offline JSON planner that preserves explicit tool choices and reports unmet requirements; it does not install, render or publish.
• An open Remotion reference project with authorised demo assets, pinned dependencies, an agent prompt, exact render commands and encoded-output checks.
• A multi-date, schema-backed integrity validator with explicit errors and independent native-export/PNG checks.

Validation: 36 tests, local retained-artifact hashes/dimensions, capability-matrix drift check, standalone package extraction/CLI smoke test and skill structure check. Six routes and eight retained exports; seventeen routing scenario specifications.

Qualification: experimental. Router remains structural_only; independent model evaluation has not run. New mixed toolchains need integration tests. The open source example has not been re-rendered on a clean machine after the lint dependency repair. Pilot time/cost was not retained. Upstream tools retain their own licences and are not bundled.

The ZIP contains only the router skill and its supporting references/scripts. Review/back up existing local skills before replacing them; no installer overwrites local work. SHA256SUMS identifies this exact archive. The tagged source contains the complete reference project.

Start here: https://github.com/markoblogo/framewright/blob/v0.1.0/docs/first-video.md
Capability matrix: https://github.com/markoblogo/framewright/blob/v0.1.0/skills/video-production-router/references/tool-matrix.md