Skip to content

chitrank2050/agent-posture

Repository files navigation

Agent Posture 🧠🏗️

Principal Grade Validated Toolchain: Native Node Security: Tier 0 Protected

Agent Posture is a high-rigor, spec-driven framework for AI Agents. It transcends standard "skills" by enforcing a 15-Pillar Constitution, multi-layer verification, and an authoritative "Skeptic's Posture." It is built for Principal Engineers who require agents to think like architects, not just coders.


🏛️ Architecture & Governance

We divide engineering rigor into three distinct layers to ensure both broad alignment and deep technical execution.

1. The Core Lens (posture-core)

The non-negotiable foundation of all engineering activity.

  • 77-Point Dossier: Atomic engineering anchors (V1-V36, S1-S41) serving as the "ground truth" for all logic.
  • Meta-Rules (The Lens): Cognitive filters for substrate-independence and official-docs-first discovery.
  • Voice Contract: A strict agreement on zero-filler communication and surgical intent.

2. Specialized Postures

Domain-specific rigor for complex systems. Each pillar is mapped to the Core Dossier but provides specialized operating rules.

  • Systems Grade (posture-backend): Atomic transactions, jittered resilience, and O(n) border safety.
  • UX Grade (posture-frontend): State-as-URL truth, hydration hygiene, and CLS-zero layout integrity.
  • Automation Grade (posture-workflows): OIDC identity, runner hardening, and SLSA 3 integrity.
  • Search Grade (posture-seo): JSON-LD schema, metadata hierarchy, and indexation hygiene.
  • Data Grade (posture-data): Migration sovereignty, tenant isolation, and deterministic seeding.
  • Zero-Trust Grade (posture-security): PII sanitization, immutable audit trails, and internal entitlements.
  • Contract Grade (posture-api): Idempotency keys, response normalization, and contract-first design.
  • Confidence Grade (posture-testing): Production-parity integration, flake-zero resilience, and edge-case focus.
  • Production-Mirror Grade (posture-infra): Health-aware orchestration, resource constraints, and rationale-driven config.
  • Communication Grade (posture-docs): Rationale-first documentation, intent-driven diagrams, and zero-stale ADRs.
  • Spec Grade (posture-readme): Professional Azure-style structure, GFM admonitions, and author signatures.
  • History Grade (posture-changelog): Conventional commit mapping, semantic sectioning, and constitutional attribution.
  • Container Grade (posture-docker): Multi-stage mandates, non-root sovereignty, and deterministic layer caching.
  • Meta Grade (posture-skill-creator): The architectural standard for authoring new engineering skills.

3. The Compiled Constitution

Our unique compiler merges core directives and 10 Architectural References into a single, high-density system prompt (dist/system-prompt.md).

  • Effect: The agent carries the full "Principal Mindset" in its active context, enabling autonomous yet governed engineering.

🛡️ Safety & Safeguards

Important

The Consent Handshake: This framework implements a "Dead Man's Switch." Agents are constitutionally forbidden from performing destructive acts (e.g., DROP TABLE, DELETE ALL) autonomously. Any Tier 0 (Black) risk requires a mandatory manual bypass.

  • Production Lock: All database operations assume a production environment by default.
  • The Halt Signal: Agents MUST HALT immediately if they detect a Tier 0 or Tier 1 anti-pattern.
  • Surgical Intent: Every contribution follows a strict anatomy: Intent → Boundary → RCA → Approach → Verification.

⚡ Technical Specification

Tip

Zero-Bloat Native Toolchain: Built for the next era of Node.js. We have removed all dependencies like commander, yargs, and ts-node.

  • Native TypeScript: Runs natively via node --experimental-strip-types.
  • Native CLI: Uses node:util/parseArgs for surgical command parsing.
  • Native ESM: Strictly uses ESM module patterns and explicit .ts extensions.
  • Native Styling: High-density terminal feedback via native ANSI escape codes.

🚀 Getting Started

Installation via registry

# Add the core engineering mind
npx skills add chitrank2050/agent-posture --skill posture-core

# Add the complete 15-Pillar constitution
npx skills add chitrank2050/agent-posture --skill \
  posture-core posture-backend posture-frontend posture-workflows \
  posture-seo posture-data posture-security posture-api \
  posture-testing posture-infra posture-docs posture-readme \
  posture-changelog posture-docker posture-skill-creator

Local Development

# Install (Zero External Dependencies)
pnpm install

# Validate all 15 skills against the Zod schema
pnpm run validate

# Compile the full constitution into a single system prompt
pnpm run compile [addon-name]

📜 Reference Library

The framework is supported by 10 high-density reference files:

  • Ref 00: The 77-Point Dossier (Atomic Anchors).
  • Ref 05: The Anti-Pattern Protocol (Hard Stops).
  • Ref 09: Calibration Diffs (Before/After logic).
  • ... and 7 others covering Stack Eras, PR Anatomy, and The Canon.

❤️ Developed by Chitrank Agnihotri

About

A high-rigor, spec-driven framework for Agentic Engineering instructions. Move beyond "skills" to a full Principal Engineering specification.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Contributors