Repository navigation
how to contribute development workflow
MyPrompt is a standalone SAFRS capsule at the repository root. Every lifecycle command runs from that directory and this capsule's own manifest, lockfile, source, Prisma schema, generated output, scripts, and tests. Nothing may require the surrounding Monorepo.
project.contract.json declares the executable stages. AGENTS.md repeats the same table.
| Stage | Program | Arguments |
|---|---|---|
| install | pnpm |
install --frozen-lockfile |
| lint | pnpm |
run lint |
| typecheck | pnpm |
run typecheck |
| test | pnpm |
run test |
| build | pnpm |
run build |
| run | pnpm |
run start |
| deployDryRun | pnpm |
run deploy:dry-run |
- Edit source.
-
pnpm run lint(eslint). -
pnpm run typecheck(tsc --noEmitafterdb:generate). -
pnpm run test(pretestrunsscripts/optimizer-acceptance.ts --dry-run, thenvitest run). -
pnpm run build(desktop:build). -
pnpm run desktop:smokefor a headless boot check. -
pnpm run deploy:dry-runto confirm the desktop artifacts exist.
pnpm run verify chains structure, lint, typecheck, test, build, smoke, and deploy dry-run.
Default risk is R2. Read .safrs/policy.json for the repository scale.
- R0 is read-only analysis.
- R1 is reversible local changes.
- R2 is boundary-affecting: desktop IPC, provider keys, dependencies, migrations, shared APIs, architecture boundaries, governance controls.
- R3 is high-impact or irreversible, such as production deployment, production credentials, financial actions, or critical clinical logic.
R2 requires designated review at the work-package or integration boundary. It does not require per-command human approval inside an already authorized package. R3 may be prepared by an agent but needs explicit human authorization before it runs.
- Provider, database, Supabase, Resend, and Xendit values live outside the repository. Never read or commit
.env. - Do not run
db:migrate,db:migrate:deploy,db:migrate:apply,db:seed, payment callbacks, or provider requests during migration verification. - Keep context isolation and IPC validation. Do not widen the preload API during ordinary changes.
- Electron is recorded as R2 in
capabilities.json; this capsule-local record mirrors the reviewed manifest.
The capsule proves independence in two halves:
- Structural:
pnpm run verify:structurerunsscripts/verify-structure.mjs, which checks required capsule entries, rejectsworkspace:or parentfile:..dependencies, and scansdesktop/,lib/,prisma/,scripts/,data/, andtypes/forD:\DEV\MonorepoorD:\Devops\abyss-monoreporeferences. - Empirical:
pnpm run verify:extractionrunsscripts/verify-extraction.mjs, which copies the capsule into a fresh temporary directory (excludingnode_modules, build output, and VCS) and runs install, structure, lint, typecheck, test, build, smoke, and deploy dry-run there. The extraction result is decisive.
See Testing, Debugging, and Tooling for the supporting detail.
Open-source prompt engineering and multi-LLM tooling by Sentra Artificial Intelligence.
MyPrompt develops practical approaches to prompt engineering, multi-LLM optimisation, reusable prompt systems, and AI-native workflows — with an emphasis on structured, interoperable, and real-world AI use.
Built in Indonesia as part of the Sentra Artificial Intelligence ecosystem.
Sentra Artificial Intelligence · Source Repository · Official Website
Dr Ferdi Iskandar — Creator & Maintainer
LinkedIn ·
ORCID ·
Hugging Face ·
Kaggle ·
Medium ·
Substack ·
X ·
Threads
MyPrompt · Sentra Artificial Intelligence · Indonesia