agentmods.dev

pixir AGENTS.md

A set of project instructions for Pixir, a local coding-agent application whose runtime manages sessions, providers, tools, and an event log.

In plain words
What is it for?
Use it when developing or modifying the Pixir coding-agent harness and when deciding which project documentation to read.
Why use it?
It gives agents the project's architecture, terminology, rules, and relevant commands before they change code.

Instructions file for CodexOpenCode

23 5d ago A scan 2,790 tokens original MIT
View the source file Ranvier-Technologies/pixir
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/ranvier-technologies/pixir/agents-md
Clone the repo
git clone --depth 1 https://github.com/Ranvier-Technologies/pixir

Made for: Codex, OpenCode.

Per session 2,790 This file is loaded in full into every session.
When invoked 2,790 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 $0.02790 $0.02790
Opus 5 $0.01395 $0.01395
Sonnet 5 $0.00558 $0.00558
Haiku 4.5 $0.00279 $0.00279

Measured 2d ago against content hash c8df2a8844bf, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

pixir 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 · 234 lines

How it starts

The opening of the file, as written. The whole thing — 234 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md - Pixir Harness

Pixir is an OTP-native, local-first coding-agent harness. The runtime spine is:

Session -> Turn -> Provider -> Tools

The append-only Log is truth, the terminal CLI is the default presenter, and ACP over stdio is the main client transport.

Progressive Discovery

Do not read the whole repo up front. Load context in this order and stop once the task is grounded enough:

  1. Start here: this file gives the architecture map, invariants, commands, and beta stance.
  2. Human quickstart: README.md and docs/open-beta-quickstart.md explain the source-install developer-preview path.
  3. Vocabulary: CONTEXT.md defines Harness, Session, Agent, Skill, Workflow, Subagent, Event, Log, History, Workspace, Tool, Host Boundary Crossing, Provider, Credential, and PATCHMD. Use those meanings exactly.
  4. Architecture decisions: read only the ADRs relevant to your change:
    • 0001 Session is the process; Agent is role/configuration.
    • 0002 ChatGPT subscription OAuth + OpenAI Responses API.
    • 0003 stateless Turns; local Log is source of truth.
    • 0004 unified Event envelope and canonical vs ephemeral events.
    • 0005 agent ergonomics: dry-run, help, structured errors, I/O discipline.
    • 0006 permission model.
    • 0007 encrypted reasoning item persistence/replay.
    • 0008 UI-agnostic Conversation driver.
    • 0009 ACP transport over stdio.
    • 0010 Agent Skills with progressive disclosure.
    • 0011 BEAM-native Subagents.
    • 0012 structural Workflows over Subagents.
    • 0013 Skill-backed Workflow Templates.
    • 0014 Workflow Checkpoint Bundles and honest partial outcomes.
    • 0015 PATCHMD customization maintenance.
    • 0016 open beta scope as source-install developer preview.
    • 0017 minimal Harness core and Presenter boundary.
    • 0018 durable History compaction and replay repair.
    • 0019 Provider usage, prompt-cache observability, and WebSocket continuation.
    • 0020 versioned Prompt Contract, cache-key family, and compaction triggers.
    • 0021 Session Resources and Image Attachments.
    • 0022 Provider-hosted Web Search evidence, not local Tool execution.
    • 0023 Skill Context Hydration as explicit, canonical, late-bound context.
    • 0024 Session Fork and Branch Summaries.
    • 0025 Hex package scope.
    • 0026 Runtime terminal-state and replay contract.
    • 0027 external command execution as a bounded host boundary.
    • 0028 Workspace Strategies and future virtual overlays.
    • 0029 Virtual Overlay changes export virtual_diff artifacts.
    • 0030 explicit apply and merge-back semantics for virtual_diff.
    • 0031 Git worktrees as lease-owned strategy for intended repo changes.
    • 0032 minimal Workflow Events for durable run decisions.
    • 0033 typed checkpoint outputs as harness-owned projections.
    • 0034 Delegate service runtime residency before async start.
    • 0035 Write-capable Sessions require an external evidence mirror.
    • 0036 Idle-timeout recovery does not auto-resume ambiguous work.
    • 0037 Anthropic provider registry routing, pa1 prompt contract, explicit cache control, and seam-parity transport.
    • 0038 Pixir Monitor is an experimental sibling source-checkout SPA using authoritative snapshots and bounded SSE invalidation hints.
    • 0039 Provider-output truncation is neutral, durable success evidence.
    • 0040 Native OpenAI compaction is a replay window on history_compaction. Overlay amendment (2026-08-16): default-on after OpenAI Responses resolve; recovery stays local. Implementation plan for #522-B/C/D: docs/design/0004-native-openai-compaction-implementation.md.
  5. Nearest local instructions: read the closest subtree AGENTS.md before editing code, docs, tests, benchmarks, or ADRs.

Read the full file on GitHub · 234 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 · 234 lines · 2,790 tokens per session scan A c8df2a8844bf

Subscribe to this mod's changes

pixir AGENTS.md is an instructions file published in the GitHub repository Ranvier-Technologies/pixir (23 stars, last pushed 5d ago), licensed MIT. It adds 2,790 tokens to every session, about $0.0140 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.

LanguagesPlanningWriting and docsacpagent-harnesscoding-agentsdeveloper-toolselixirotp
Related

Other instructions, from other repositories

raxol CLAUDE.md

Instructions for DROOdotFOO/raxol, covering claude.md, essential commands, initial setup, building & compilation and testing.

DROOdotFOO/raxol · 17,878 tokens

RepoResident AGENTS.md

Instructions for ychamel/RepoResident: Agent entry point: read CLAUDE.md in full — it is the operating manual — then follow its Session protocol. Everything else (state, workflows, project knowledge) is routed from there via the .agent/ directory.

ychamel/RepoResident · 50 tokens

agent-harness-skills AGENTS.md

Instructions for yfge/agent-harness-skills, covering agent harness skills - repository instructions, purpose, rules, validation and commit discipline.

yfge/agent-harness-skills · 450 tokens

agent-harness-skills GEMINI.md

Instructions for yfge/agent-harness-skills: This extension provides reusable skills for repository harness design.

yfge/agent-harness-skills · 111 tokens

pi-dsh AGENTS.md

AGENTS.md instructions for abhishekgahlot2/pi-dsh, covering pi-dsh, the contract (read before changing anything), layout, commands and vendoring policy ("maintain both forks").

abhishekgahlot2/pi-dsh · 1,355 tokens

pi-dsh CLAUDE.md

Claude Code instructions for abhishekgahlot2/pi-dsh: All project instructions live in AGENTS.md — read it first and follow it exactly.

abhishekgahlot2/pi-dsh · 88 tokens