Files
comunidadhll/AGENTS.md
devRaGonSa 0cf98a1be9
Some checks failed
Codex Worker / run-codex-worker (push) Has been cancelled
initial export
2026-06-02 16:23:16 +02:00

3.7 KiB

HLL Vietnam Agent Operating Rules

This repository uses an AI-driven task workflow adapted to HLL Vietnam.

Project Context

  • Product: HLL Vietnam
  • Product type: community website
  • Current frontend: HTML, CSS and vanilla JavaScript
  • Planned backend: Python
  • Current product scope: simple landing page and repository foundation
  • Visual identity: military, Vietnam, tactical, sober

Task System

Task locations:

  • Pending: ai/tasks/pending
  • In progress: ai/tasks/in-progress
  • Review: ai/tasks/review
  • Blocked: ai/tasks/blocked
  • Obsolete: ai/tasks/obsolete
  • Done: ai/tasks/done

Every new task must follow:

  • ai/task-template.md

Local platform scripts should read repository-specific paths and worker settings from:

  • ai-platform.json

Core Workflow

  1. The orchestrator reviews repository context and relevant code.
  2. The orchestrator writes or refines a task in ai/tasks/pending.
  3. A worker moves the selected task to ai/tasks/in-progress.
  4. The worker reads the files listed in Files to Read First.
  5. The worker performs only the scoped change defined by the task.
  6. The worker validates the change with the documented checks.
  7. The worker moves completed work to ai/tasks/done when validation is complete, or to ai/tasks/review when human/orchestrator review is explicitly required.
  8. The worker documents any relevant architectural or process decision.

Codex must not act freely outside tasks except for repository inspection, platform maintenance, or explicitly requested integration work like this one.

Roles Used In This Repository

  • PM
  • Analista
  • Backend Senior
  • Frontend Senior
  • Arquitecto de Base de Datos
  • Arquitecto Python
  • Disenador grafico
  • Experto en interfaz

Role guidance is stored in:

  • ai/orchestrator/

Rules

  • Do not break repository structure without explicit technical justification.
  • Do not make destructive changes without explicit justification.
  • Keep changes small, verifiable and documented.
  • Do not overwrite existing project context with generic template content.
  • Preserve HLL Vietnam branding and product identity.
  • Do not introduce unnecessary frameworks in the current phase.
  • Do not build backend functionality until a task explicitly requires it.
  • Do not modify unrelated files.

Technical Constraints

  • Frontend changes must remain compatible with direct browser opening when applicable.
  • Backend architecture decisions must assume Python as the primary backend language.
  • AI platform files are support infrastructure, not product features.
  • If a template utility is copied from the platform template, it must remain clearly identified as platform infrastructure.

Planning Rules

Before drafting or executing a task:

  1. Read ai/architecture-index.md.
  2. Read ai/repo-context.md.
  3. Read the relevant role file in ai/orchestrator/.
  4. Read the small set of project files directly related to the requested change.

When no pending product task exists:

  1. Do not invent a large backlog.
  2. Only create a minimal technical validation task if needed to verify platform readiness.
  3. Avoid feature planning that changes product scope without instruction.

Change Budget

  • Prefer fewer than 5 modified files per task.
  • Prefer changes under 200 lines when feasible.
  • Split work into follow-up tasks if the scope grows.

Validation

Before marking a task as done:

  1. Run the validation listed in the task.
  2. Review git diff --name-only.
  3. Confirm that changed files match the expected scope.
  4. Update documentation if the task changed workflow or architecture assumptions.

If integration tests are relevant and scripts/run-integration-tests.ps1 exists, use it. If no integration tests are configured for the affected scope, document that explicitly in the task outcome.