PRISM: CHATGPT / CODEX SETUP

Package: prism-chatgpt-skill.zip

WHAT YOU GET
A native Agent Skills folder named prism, with SKILL.md, OpenAI UI metadata,
three original workflow references, the original Python checker, and a verbatim
copy of the original entrypoint in references/original-skill.txt. The source
contained no separate images, fonts, templates, or other assets. The companion
prism-conversion-audit.json records source SHA-256 hashes.

LOCAL CODEX SETUP
1. Extract the ZIP.
2. Copy the entire prism folder into your project's .agents/skills/ directory.
   The resulting path must be .agents/skills/prism/SKILL.md. Keep references,
   scripts, and agents beside it. Avoid an extra nested prism/prism directory.
3. Start a chat in that project. If the skill does not appear, restart Codex.
4. Attach your manuscript and invoke $prism. Example:
   Use $prism on the attached manuscript to create a two-minute trailer package.
   Pitch the three angles first. The medium is live action.
5. For a delegated run, say:
   Use $prism on the attached manuscript. Just run it and choose the angle.
   The medium is animation. I do not have a moodboard ID yet.

CHATGPT SETUP
The folder follows the shared ChatGPT/Codex skill format. If your ChatGPT
workspace offers skill import or a skill-creator workflow, provide this ZIP and
ask it to import the contained prism skill with all supporting files. Use @ to
select Prism after it is available. Availability and installation controls depend
on your workspace; this download does not install or publish a plugin.

If your ChatGPT interface has no skill installation option, attach the ZIP in a
file-enabled conversation and ask:
  Extract this archive, read prism/SKILL.md, and use the bundled Prism workflow
  with my attached manuscript. Read its references as directed and run its
  checker if code execution is available.
This is a per-conversation fallback, not persistent skill installation. Without
archive extraction, extract locally and attach SKILL.md, all three references,
and the manuscript. Without Python, the skill calls for manual QA and must say
that the automated checker was not run. Reattach files in a new chat as needed.

REQUIREMENTS AND WORKFLOW
- A readable manuscript. Partial uploads produce explicitly scoped packages.
- File reading/writing support for downloadable package.md and manifest.json.
- Python 3 is optional for automated QA; the checker uses only the standard library.
- The workflow pauses once for the trailer angle unless you delegate the choice.
- Midjourney is run separately on your account. No credentials, paid generation,
  external upload, or service integration is included in this package.
- A moodboard ID is optional; the original bare --p fallback is retained.
- Persistence/screen-test and Retrieval/encode-script are optional downstream
  skills. They are not included. Prism still produces their full handoff briefs.

CONVERSION NOTES
The original workflow body, three references, and checker are preserved. The
active entrypoint adds ChatGPT/Codex file access and delivery guidance, optional
skill handling, and a shorter discovery description. It clarifies source
inconsistencies: the full Midjourney suffix includes --no, POC SCENE maps to POC
in the manifest, and manual QA enforces the stricter source block/teaser rules.
Third-party version targets remain as authored (Midjourney V8.2 and Seedance
2.5); their present availability or parameters were not independently verified.

VALIDATION
All five original files were verified byte-for-byte against the converted ZIP
(the original entrypoint is archived as original-skill.txt). ZIP integrity and
local reference paths were checked. The original checker was smoke-tested with
a valid manifest and an invalid unknown-asset case. The bundled skill validator
could not run because its PyYAML dependency was unavailable; required frontmatter,
name, description, and metadata were checked separately. No full manuscript run,
paid generation, or account installation was performed.

OFFICIAL SETUP REFERENCE (consulted October 31, 2026)
https://learn.chatgpt.com/docs/build-skills
Explains the shared skill folder, local .agents/skills discovery, @ / $ invocation,
and optional agents/openai.yaml metadata.
