orcbot AGENTS.md

A repository guide for OrcBot, a TypeScript command-line AI agent that plans tasks, calls tools, and keeps its state in files. It explains the main commands and the roles of the system's core layers.

In plain words
What is it for?
Building and testing OrcBot, locating architecture components, and changing its planning, tool-use, parsing, or safety behavior.
Why use it?
It helps contributors understand where the command-line interface, agent loop, decision logic, and tool safeguards are implemented.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/fredabila/orcbot/agents-md
Clone the repo
git clone --depth 1 https://github.com/fredabila/orcbot

Made for: Codex, OpenCode.

Per session 813 This file is loaded in full into every session.
When invoked 813 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00813 $0.00813
Opus 5 $0.00407 $0.00407
Sonnet 5 $0.00163 $0.00163
Haiku 4.5 $0.00081 $0.00081

Measured 2d ago against content hash 52dd99d344ce, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

orcbot AGENTS.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 2d 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.

AGENTS.md · 47 lines

How it starts

The opening of the file, as written. The whole thing — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

This file provides guidance to WARP (warp.dev) when working with code in this repository.

Common commands

  • Install deps: npm install
  • Build (tsc): npm run build
  • Fast build: npm run build:fast
  • Dev CLI (ts-node): npm run dev
  • Run built CLI: npm run start (runs dist/cli/index.js)
  • Tests (Vitest): npm test
  • Watch tests: npm run test:watch
  • Run a single test file: npx vitest run tests/<file>.test.ts
  • Run a single test by name: npx vitest run -t "test name"
  • Browser tooling smoke test: npm run browser:test
  • Lint: no lint script is defined in package.json

High-level architecture (big picture)

  • CLI entrypoint: src/cli/index.ts wires the TUI, CLI commands, and bootstraps subsystems.
  • Agent core loop: src/core/Agent.ts orchestrates the action loop (simulate → decide → execute tools → memory → termination review).
  • Decision stack:
    • src/core/DecisionEngine.ts builds prompts and calls the LLM.
    • src/core/ParserLayer.ts normalizes LLM output to structured JSON with fallbacks.
    • src/core/DecisionPipeline.ts applies guardrails on tool calls (dedup, loop detection, safety checks). Guardrails are intentional—avoid weakening them.
    • src/core/SimulationEngine.ts creates a pre-plan before execution starts.
    • src/core/prompts/ contains modular helpers activated by PromptRouter.
  • Memory system (critical):
    • src/memory/MemoryManager.ts handles short/episodic/long memory and consolidation.
    • src/memory/ActionQueue.ts is the durable priority queue for tasks (retry, TTL, chaining).
    • Storage is file-backed via src/storage/JSONAdapter.ts (atomic writes + backups).
    • Vector memory (src/memory/VectorMemory.ts) is optional and file-backed.
    • Important constraints from existing instructions: keep saveMemory content short (<500 chars), include metadata (actionId, step, skill) for step memories, and avoid storing secrets in memory.
  • Skills system:
    • Core skills registered in src/core/Agent.ts.
    • Dynamic plugins loaded by src/core/SkillsManager.ts from ~/.orcbot/plugins/ (or ./plugins).
  • Channels: src/channels/ implements Telegram (Telegraf), WhatsApp (Baileys), Discord (discord.js), and Gateway (Express+WS). Inbound messages write short memory and push tasks to the queue.
  • Web/browser tooling: src/tools/WebBrowser.ts wraps Playwright and provides search fallbacks (Serper → Google → Bing → DuckDuckGo).
  • Config + runtime tuning:
    • src/config/ConfigManager.ts loads orcbot.config.yaml with hot-reload and feature toggles.
    • src/core/RuntimeTuner.ts adjusts runtime limits based on signals.
    • src/core/MultiLLM.ts routes providers by model prefix and falls back on errors.
  • Data paths: runtime state lives under ~/.orcbot/ (or ORCBOT_DATA_DIR) including config, memory, queue, logs, profiles, and plugins.

Read the full file on GitHub · 47 lines

Changes

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.

  1. 2d ago First seen · 47 lines · 813 tokens per session scan A 52dd99d344ce

Subscribe to this mod's changes

orcbot AGENTS.md is an instructions file published in the GitHub repository fredabila/orcbot (11 stars, last pushed 5mo ago), licensed MIT. It adds 813 tokens to every session, about $0.0041 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.