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/defkode/orbx/agents-mdgit clone --depth 1 https://github.com/defkode/orbxWhat 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 | $0.01549 | $0.01549 |
| Opus 5 | $0.00775 | $0.00775 |
| Sonnet 5 | $0.00310 | $0.00310 |
| Haiku 4.5 | $0.00155 | $0.00155 |
Grade A, and why
orbx 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 yesterday.
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 — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file — also read as CLAUDE.md (a symlink to it) — provides guidance to coding agents (Claude Code, GitHub Copilot CLI, and others) when working with code in this repository.
What this is
orbx (Orbstack linuX) is a single-file, zero-dependency Bash CLI that
spins up per-project OrbStack Linux machines. It derives
a machine name from $PWD, mounts the project into the VM, provisions it from a
cloud-init template, waits for readiness, and shells in. All product code lives
in one script: bin/orbx.
Commands
bats test/ # run all tests
bats test/shellrun.bats # run one file
bats -f "run forwards flags" test/ # run tests matching a name
shellcheck bin/orbx test/helpers/orb-stub script/release # lint (CI pins v0.11.0)
script/release X.Y.Z # cut a release (see below)
There is no build step — bin/orbx is the shipped artifact.
Architecture
Dual-mode script. bin/orbx runs as an executable or is sourced. The
guard at the bottom (if [[ "${BASH_SOURCE[0]}" == "${0}" ]]) means main
runs only when executed; when sourced, only the orbx::* functions load. Tests
rely on this: orbx_source (unit-test individual functions) vs orbx_run
(integration through main). set -euo pipefail is set inside that guard, so
it does not apply when sourced.
Config layering. Precedence, most-specific wins:
CLI flags > ./.orbxrc > ~/.orbx/config > built-in defaults.
orbx::effective_config merges the two flat key = value files over defaults;
orbx::resolve_context then re-reads that plus the parsed ORBX_FLAG_* vars and
populates the ORBX_* run-state globals every command reads. Config is parsed
line-by-line (orbx::parse_config_file) and never sourced or eval'd — a
config file can't execute code. Keep it that way. Values can't contain #
(everything after # is a comment).
Template resolution. Templates are cloud-init YAML. A bare name is resolved
through a search path (orbx::resolve_template): $ORBX_TEMPLATE_DIR →
~/.orbx/templates → bundled dir → <script>/../templates. A user file shadows
the bundled one of the same name. A path-shaped value — one with an explicit
./, ../, / or ~/ prefix, or a .yaml/.yml extension — skips the search
path entirely and is used as given, resolved against $PWD (the same anchor as
./.orbxrc). That is what lets a repo commit its own template alongside its
config; the two forms are disjoint, so a bare name never changes meaning.
ORBX_BUNDLED_TEMPLATE_DIR is the literal @TEMPLATE_DIR@ placeholder in-repo;
the Homebrew formula rewrites it at install time (unstamped in a git checkout,
which is why the <script>/../templates fallback exists).
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.
- yesterday First seen · 104 lines · 1,549 tokens per session scan A d3c86fde95e0
orbx AGENTS.md is an instructions file published in the GitHub repository defkode/orbx (2 stars, last pushed 2d ago), licensed MIT. It adds 1,549 tokens to every session, about $0.0077 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
mirage CLAUDE.md
Instructions for strukto-ai/mirage, covering claude.md, repo layout, typescript packages, python/typescript parity and module layout.
flox AGENTS.md
AGENTS.md instructions for flox/flox, covering claude.md, project overview, development setup, already in dev shell (innixshell is set) — run directly and not in dev shell — wrap with nix develop -c.
coder CLAUDE.md
Claude Code instructions for coder/coder, a project described as: Secure environments for developers and their agents.
lade AGENTS.md
Instructions for zifeo/lade, covering agents.md, build & test, house rules, project layout and maintainer note.
senior-engineering-partner CLAUDE.md
Instructions for bjgreenberg/senior-engineering-partner, covering claude.md — senior-engineering-partner, model requirement (read before editing anything), repo shape, pr rules and gates & sharp edges.
flox CLAUDE.md
Claude Code instructions for flox/flox, a project described as: The Deterministic Foundation for your SDLC.