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/bkataru/powerglide/agents-mdgit clone --depth 1 https://github.com/bkataru/powerglideWhat 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 | $0.01089 | $0.01089 |
| Opus 5 | $0.00544 | $0.00544 |
| Sonnet 5 | $0.00218 | $0.00218 |
| Haiku 4.5 | $0.00109 | $0.00109 |
Grade A, and why
powerglide 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 3d 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 — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.
powerglide Agent Protocol & Integration Guide
Overview
powerglide is designed to be both a standalone CLI tool and a high-performance substrate for other agents (like Claude Code, oh-my-opencode, etc.) to build upon. This guide details the Ralph Loop protocol, PTY management, and how to integrate with powerglide programmatically.
The Ralph Loop Protocol
Every session follows an explicit 11-state transition model. Agents MUST NOT skip states or exit implicitly.
| State | Purpose | Transition Signal |
|---|---|---|
idle |
Waiting for initialization | Start session |
load_tasks |
Fetching from task-queue.json |
Queue not empty |
pick_task |
Selecting highest priority task | Task assigned |
thinking |
LLM cognition / completion | LLM response received |
tool_call |
Parsing tool use blocks | Tools ready |
executing |
Subprocess in isolated PTY | Tool finished |
observing |
Result aggregation / analysis | Next action decided |
verify |
Running automated checks | Success/Failure |
commit |
State persistence | Task status updated |
done |
Clean termination | <POWERGLIDE_DONE> |
failed |
Unrecoverable error | <POWERGLIDE_ERROR> |
Termination Signals
- Success: The session MUST output
<POWERGLIDE_DONE>to stdout upon completion of all tasks. - Failure: The session MUST output
<POWERGLIDE_ERROR>followed by a diagnostic message.
Velocity Control
Velocity is a floating-point multiplier (f64) on a 1000ms base delay.
- Formula:
delay_ms = 1000 / velocity - Default:
1.0(1000ms) - Speeding up:
2.0(500ms),4.0(250ms) - Slowing down:
0.5(2000ms),0.25(4000ms)
CLI flags: --velocity accepts floating-point values.
Agents can self-throttle via CLI flag or config:
powerglide run --velocity 0.5 "task" or set velocity in ~/.config/powerglide/config.json
Reliable PTY Execution
powerglide runs all tools in a pseudoterminal (PTY). This ensures:
- Interactive behavior: Tools like
gitornpmbehave as if they are in a real terminal. - ANSI colors: preserved in logs.
- Exit Code capture: Reliability via
waitpidwithWNOHANGpolling and/proc/<pid>/statusfallback.
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.
- 3d ago First seen · 112 lines · 1,089 tokens per session scan A 4f2d6ca39c24
powerglide AGENTS.md is an instructions file published in the GitHub repository bkataru/powerglide (6 stars, last pushed 6mo ago), licensed MIT. It adds 1,089 tokens to every session, about $0.0054 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 2026-08-31.
Other instructions, from other repositories
Arkana CLAUDE.md
Claude Code instructions for JameZUK/Arkana, covering arkana development guide, what is arkana?, project structure, running tests and unit tests (no server needed, 2 seconds).
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).