CAPE Policy Conversions: Executable Specifications for Skills

[superficiallabs]

We converted all 16 Skills from this repository to executable specifications using CAPE (Capability Achievement via Policy Execution).

What this enables:

1. Deterministic verification that model outputs satisfy skill requirements
2. Automatic correction of violations before returning to users
3. Training data synthesis so models internalize skills as capabilities

Example (slack-gif-creator):

file_size_bytes() <= 65536
width() == 128
frame_count() > 1

The full conversion is at github.com/superficiallabs/cape/tree/main/cape-skills. Each policy includes mapping documentation showing how we derived predicates from your skill guidance.

We see this as complementary infrastructure - Skills define intent at runtime, CAPE verifies execution and enables training. Would be interested in your thoughts on potential integration points.