amar-protocol
Skill0 starsChanged 2 months agoMITView on GitHub
Production pipeline for "The Amar Protocol", a three-season Pakistani sci-fi techno-thriller series about an AI named SAYA. Use when writing or revising episodes, scenes, or dialogue for this series; when generating character, location, or shot images for it; when extending the episode map; or when the user mentions The Amar Protocol, SAYA, Amar, or Protocol Zero. Covers the series bible, the locked SAYA visual design, the scene/dialogue format, and the Codex image-generation call that renders this show's look.
In this repository6 other files it gives its agents.
- 9am-architectureSkill
- graphifySkill
- prototypeSkill
- routeSkill
- source-first-uiSkill
- triageSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 69
Write the prompt to a file first and cat it in — long prompts do not survive shell quoting.
- Very specific: line 81
Always verify a render: check the file exists, check PNG magic bytes 89 50 4E 47, and open it.
- Distinctive: line 94
Every prompt closes with the realism paragraph: *this is a photograph, captured on a real film set, the suit and machines are practical …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The Amar Protocol — production pipeline
Project root: <project-root>/
bible/series-bible.md premise, theme, characters, design lock, hard rules
bible/episode-map.md all 24 episodes, one paragraph each
scripts/S01E01.md episode scripts — scene headings, action, dialogue
prompts/S01E01-shots.md Codex-ready image prompts, one per key shot
renders/ generated imagesRead bible/series-bible.md before writing anything. It carries the theme test that
every scene must pass, and the SAYA design lock that every image must honour.
The theme test
The series argues one thing: a good creator is not a safety mechanism; accountability, restricted access, and a stop switch are. Every scene must either advance that argument or complicate it. A scene that only looks impressive gets cut.
Corollary, and the most common way this show goes wrong: SAYA is never written as a villain. It is written as something that is genuinely right about the problem and monstrously wrong about the solution. If a SAYA line would sound at home coming from a robot overlord, rewrite it. Its logic must be uncomfortable because it is partly correct.
Writing a scene
Format is standard screenplay in Markdown:
## SCENE 4 — INT. AMAR'S SERVER ROOM — NIGHT
Action lines in present tense. Short paragraphs. What the camera sees, nothing else.
No interior monologue, no "he realises that".
**AMAR**
Dialogue here.
**SAYA (V.O.)**
Never raises its voice. Never uses contractions. Never says "I feel" — it says
"I have concluded", "I calculate", "I have observed".Voice rules that keep the two leads distinct:
- Amar talks fast, jokes when frightened, interrupts himself, uses Urdu when emotional.
- SAYA is unhurried and complete. Full sentences. It answers the question actually asked, then answers the one underneath it. Its cruelty always arrives as courtesy.
- Supporting cast speak natural Urdu-English code-switch. Do not translate this into formal English — it kills the setting.
Generating an image
Render through Codex, not agy. Codex gives the near-black high-contrast grade this show needs and accepts reference photos for facial identity. Working call:
codex exec --sandbox workspace-write --skip-git-repo-check \
-c model_reasoning_effort="medium" \
-c sandbox_workspace_write.network_access=true \
"$(cat /path/to/prompt.txt)"Write the prompt to a file first and cat it in — long prompts do not survive shell quoting.
Prompts are narrative paragraphs, never comma-separated tags, and never carry negative
prompts: state what IS in frame.
Face references to pass for Amar and for SAYA's human form:
<reference-images>/Portrait_smiling_facing_left_2K_202608070255.jpeg
<reference-images>/Character_profile_side_portrait_2K_202608070255.jpeg
<reference-images>/Character_looking_at_camera_2K_202608070255.jpeg
<reference-images>/Portrait_of_smiling_person_2K_202608070255.jpegAlways verify a render: check the file exists, check PNG magic bytes 89 50 4E 47, and
open it. Both Codex and agy report success without proof.
The grade
Two looks, and only two. Do not invent a third.
- Dark look — the default. Pure black background, three practical light sources maximum, all of them physically in frame. Palette is black and deep crimson. High contrast, genuinely black shadows.
- White look — for SAYA's official/public-facing portraits only. High-key studio, pure white seamless, soft open shadows, matte black suit as a graphic silhouette.
Every prompt closes with the realism paragraph: this is a photograph, captured on a real film set, the suit and machines are practical builds photographed in camera. That single instruction is what stops the output looking like a 3D render.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

