Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/youssefsiam38/agentpg/claude-mdgit clone --depth 1 https://github.com/youssefsiam38/agentpgWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/youssefsiam38/agentpg/claude-md)<a href="https://agentmods.dev/instructions/youssefsiam38/agentpg/claude-md"><img src="https://agentmods.dev/badge/instructions/youssefsiam38/agentpg/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.11693 | $0.11693 |
| Opus 5 | $0.05847 | $0.05847 |
| Sonnet 5 | $0.02339 | $0.02339 |
| Haiku 4.5 | $0.01169 | $0.01169 |
Grade A, and why
agentpg CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 1,285 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AgentPG - Event-Driven AI Agent Framework
Event-driven Go framework for building async AI agents using PostgreSQL for state management and distribution. Provides stateful, distributed, and transaction-safe agent execution with Claude Batch API integration.
Architecture Overview
AgentPG Architecture
Layer 1: HTTP Server Layer (Optional)
- ui.UIHandler() - /ui/* (HTMX + SSR)
- Components: Dashboard, Sessions/Runs, Chat Interface, Monitoring
Layer 2: Workers
- Client 1 (Worker), Client 2 (Worker), Client N (Worker)
- Deployed as k8s pods
Layer 3: PostgreSQL Database
- Tables: Sessions, Runs, Iterations, Tools, Messages, Instances, Agents
- LISTEN/NOTIFY channels:
- agentpg_run_created
- agentpg_tool_pending
- agentpg_run_state
- agentpg_tools_complete
Layer 4: Claude Batch API
- 24h processing window
- Async submission/polling
Key Design Principles
- PostgreSQL as Single Source of Truth: All state stored in PostgreSQL. No in-memory state that could be lost.
- Event-Driven with Polling Fallback: Uses LISTEN/NOTIFY for real-time events, with polling as fallback.
- Race-Safe Distribution: Uses
SELECT FOR UPDATE SKIP LOCKEDfor safe work claiming across workers. - Transaction-First:
RunTx()accepts user transactions for atomic operations. - Database-Driven Agents: Agents are database entities with UUID primary keys, not per-client registrations. Tools are registered per-client.
- Multi-Level Agent Hierarchies: Agents can be tools for other agents (PM → Lead → Worker pattern).
Core Concepts
Client
A Client represents a single worker instance. Multiple clients can run across different processes/pods, sharing work from the same database.
client, err := agentpg.NewClient(driver, &agentpg.ClientConfig{
APIKey: os.Getenv("ANTHROPIC_API_KEY"),
Name: "worker-1",
})
Agent
An Agent is a Claude-powered AI with a specific role, model, and system prompt. Agents are database entities identified by UUID, created via CreateAgent() after the client starts.
agent, err := client.CreateAgent(ctx, &agentpg.AgentDefinition{
Name: "assistant",
Model: "claude-sonnet-4-5-20250929",
SystemPrompt: "You are a helpful assistant.",
})
// Use agent.ID (uuid.UUID) when running
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 1,285 lines · 11,693 tokens per session scan A 80bcee10c64c
agentpg CLAUDE.md is an instructions file published in the GitHub repository youssefsiam38/agentpg (11 stars, last pushed 5mo ago), licensed MPL-2.0. It adds 11,693 tokens to every session, about $0.0585 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
supabase copilot-instructions.md
Copilot instructions for supabase/supabase, covering copilot code review instructions, review policy — read this first, confidence threshold, what not to comment on and what to comment on (priority order).
prest copilot-instructions.md
Copilot instructions for prest/prest, covering copilot instructions for prest go backend, project scope, core engineering principles, solid principles for go and s — single responsibility.
postgres-skills AGENTS.md
Instructions for neondatabase/postgres-skills, covering agents.md, repository overview, creating a new skill, directory structure and naming conventions.
pg_exporter AGENTS.md
Instructions for nbari/pg_exporter, covering agents.md, project shape, first steps, devpod workflow (preferred) and local commands.
counterpoise-ledger CLAUDE.md
Instructions for jeffjjohnston/counterpoise-ledger, covering claude.md, project overview, development commands, essential commands and testing.
database-mcp CLAUDE.md
Instructions for hadi21k/database-mcp, covering claude.md, project overview, commands, architecture and layers.