geminify

{
  "name": "geminify",
  "description": "Two modes. Mode A adds a Gemini-calibrated `gemini.md` to an existing skill, so the same skill works on Google's models as well as Claude's: reads the target, scans it for the unbounded scopes and assumed verification that Gemini handles differently, assembles the file from a shared core plus only the modules the skill's own subject matter earns, tags every claim by how strong its evidence actually is, gates the result by checking each quoted vendor sentence appears verbatim in Google's published guidance, then installs the conditional pointer and bumps the version. Mode B calibrates the session the runner is already in, when a Gemini model is serving it and no target is named: `session_calibrate.py` reports which loaded skills ship a `gemini.md`, which source names have an invalid namespace or are unpublished by their marketplace and may fail at runtime, and which referral lanes are out-of-family for the serving model, then the runner posts a receipt ledger where a Skill tool call discharges a row and conforming output does not. Use whenever someone wants a skill to work on Gemini — \"geminify design-craft\", \"add Gemini support to this skill\", \"make this skill work in Antigravity\", \"write a gemini.md for X\", \"/geminify:geminify\" — whenever a skill produced weak output under a Gemini model and the skill itself is not at fault, and whenever a session is served by a Gemini-family model and should apply the overrides to the skills it has loaded. NOT for fixing a skill that is broken for every model (use improve-skill:improve-skill), and NOT for authoring a new skill (use create-skill:create-skill).",
  "version": "0.6.1",
  "author": {
    "name": "Fledgeling",
    "email": "[email protected]"
  }
}