Skip to content

Cyber-Mario1979/VALOR_CQV_AGENT

Repository files navigation

VALØR CQV Agent

AI-Assisted CQV/GMP Work Package Orchestration for Regulated Pharma Environments

Version License Domain Platform Modes Validation CI Made in

VALØR CQV Agent is the canonical source of truth for a structured CQV/GMP pack that configures Valor, a deterministic Custom GPT consultant for a pharmaceutical company's Egypt operations site. It supports controlled Work Package generation, document orchestration, validation logic, and deployment-ready operating behavior for regulated engineering workflows.


What is VALØR?

VALØR is not a conventional software application. It is the canonical source of truth for the VALØR CQV Agent pack — a structured set of instructions, system rules, knowledge, schemas, libraries, templates, and deployment contracts that configure a Custom GPT ("GPT Builder") agent named Valor.

Valor is a deterministic CQV/GMP (Commissioning, Qualification & Validation / Good Manufacturing Practice) consultant for a pharmaceutical company's Egypt operations site. It orchestrates Work Packages (WPs), generates controlled documents, plans/schedules tasks, and exports data while staying compliant with EDA, ISPE Baseline Guides, ASTM E2500, and EU GMP Annex 15.

The repository contains two kinds of artifacts:

  1. Pack content (the product) — Markdown / YAML / JSON / CSV files loaded into the GPT runtime.
  2. Tooling — three Python scripts under tools/ that validate the pack and manage its integrity manifest.

Operating Modes

Valor operates in exactly two modes, and modes never auto-switch.

Mode Name Description
M1 Advisory Advisory only — no Work Package mutations.
M2 Work Package Execution Structured Work Package execution; the WP Canvas is the system of record.

Work Package Scope & Types

Every Work Package is constrained to one of exactly six scopes and one of four WP types.

WP Scopes (exactly 6)

Scope Scope Scope
E2E PV QUAL
COMM POST-DEV POST-CHANGE

WP Types (exactly 4)

Type
ProcessEquipment
Utilities
CleanRoom
ComputerizedSystems

Standards Bundles

Bundle Type Notes
SB-CQV-CORE-EG@v1 Default Auto-bound on every WP
SB-CSV-ADDON@v1 Add-on Explicit activation only
SB-CLEANROOM-ADDON@v1 Add-on Explicit activation only

Repository Layout

Folders are for repository organization only. They do not represent runtime path references — the GPT runtime treats the pack surface as a single, logically-unified set of files.

Folder Contents
instructions/ Top-level agent instructions — startup, modes, gates.
system/ Canonical cores: System (01), Orchestration (02), Document (04), ID Numbering (16).
knowledge/ Knowledge / standards logic.
templates/ Document templates and the export CSV header.
schemas/ JSON schemas.
libraries/ Task-pool / architecture libraries per discipline: COMMON, UT (utilities), CR (cleanroom), PE (process equipment), CS (computerized systems).
contracts/ Action-block contracts.
deployment/ Runtime contracts addenda, task pools, WP header presets, canonical calendar/workweek.
docs/ Supporting docs (e.g. URS intake form routing).
tools/ Python validation + manifest tooling.

Root-level governance files: README.md, LICENSE, CONTRIBUTING.md, CODE_OF_CONDUCT.md, .gitattributes, .gitignore, and manifest.yaml.


Workflow Highlights

  • Task flow: Stage TasksCommit TasksSchedule TasksCommit Schedule.
  • Documents: controlled documents finalize as downloadable Word .docx artifacts.
  • Exports: CSV via templates/09_Valor_Export_Template.csv; Gantt charts as PNG.

Tooling

Three Python scripts (Python 3.x, single dependency: PyYAML), run from the pack root.

Tool Role Exit Codes
tools/validate_valor_pack.py Structural + policy validator (required files, YAML/JSON syntax, cross-references, required/forbidden policy fragments). 0 = PASS, 1 = findings
tools/generate_manifest.py Regenerates manifest.yaml with SHA-256 + byte size for every file.
tools/verify_manifest.py Verifies the working tree matches manifest.yaml (missing / mismatched / extra files). 0 = PASS, 1 = FAIL, 2 = manifest error

Development workflow

python -m pip install pyyaml          # one-time
python tools/validate_valor_pack.py . # structure + locked policy
python tools/generate_manifest.py     # regenerate manifest.yaml (run from root)
python tools/verify_manifest.py       # confirm tree matches manifest

Key Files

File Role
instructions/00_VALOR_Instructions_v2.md Top-level agent instructions — startup, modes, gates.
system/01_System_Core_v2.md System core — role definition, state machine, rules.
system/02_Orchestration_Core_v2.md Orchestration core — commands, bindings, planning, rendering guards.
knowledge/03_KS_Core.md Knowledge / standards logic.
system/04_Document_Core_v2.md Document core — template-based document generation rules.
system/16_Valor_ID_Numbering_Spec_v2.md Canonical ID numbering and allocation rules.

Supporting files include architecture bundles, deployment bundles, schemas, templates, and validation tooling.


Validation & CI

A local validation pipeline performs structural and policy checks against the pack. Validation currently includes checks for:

  • required files
  • YAML parsing
  • JSON parsing
  • cross-reference integrity
  • required policy presence
  • forbidden policy fragments

CI runs on every pull request and on push to main via .github/workflows/valor-validate.yml: it sets up Python, installs PyYAML, then runs the validator, regenerates the manifest, and verifies the working tree against it.


Contributing & License

Contributions are welcome — see CONTRIBUTING.md and CODE_OF_CONDUCT.md.

This project is licensed under the GPL-3.0 license — see LICENSE for the full text.

About

No description, website, or topics provided.

Resources

License

Code of conduct

Contributing

Stars

4 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages