controller-behavior-test

A command-line test harness for checking how code-orchestrating hosts behave in scripted scenarios. It currently covers Codex with bash and includes paths for Claude Code and multiple hosts.

In plain words
What is it for?
Use it to run the `doctor-loads` scenario, test Codex or Claude Code host runners, and run registered scenarios across available orchestrators.
Why use it?
It helps catch changes that make an orchestrator behave differently across hosts or communication methods. Test transcripts make the observed behavior available for review.

Command

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 commands/simonrowland/goal-flight/controller-behavior-test
Clone the repo
git clone --depth 1 https://github.com/simonrowland/goal-flight
Per session 14 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,781 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00014 $0.01781
Opus 5 $0.00007 $0.00890
Sonnet 5 $0.00003 $0.00356
Haiku 4.5 $0.00001 $0.00178

Measured yesterday against content hash 30df848a6daa, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

controller-behavior-test 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.

commands/controller-behavior-test.md · 143 lines

How it starts

The opening of the file, as written. The whole thing — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.

controller-behavior-test [--controller codex] [--scenario doctor-loads]

Read:

  • docs-private/plans/controller-behavior-harness-plan.md (when present locally)

Runs the file-backed orchestrator behavior harness. Wave 1 implements Codex + bash-tail with the doctor-loads scenario. The Claude Code regression path uses claude-code-cli-acp through Goal Flight ACP so it stays on the subscription-routed interactive shim. The multi-host live tier enumerates available orchestrators from scripts/hosts/controller/probe_matrix.py, filters them through GOALFLIGHT_LIVE_CONTROLLERS, and runs every registered scenario against that intersection.

Host runners

Orchestrator Transport Bash test Transcript path
codex bash-tail codex exec tests/bash/test-controller-behavior-codex.sh harness temp tail
claude-acp ACP shim via claude-code-cli-acp tests/bash/test-controller-behavior-claude-code-acp.sh docs-private/reviews/<date>-chunk-15/<scenario>.transcript.log
env-gated available set host runner from behavior_scenario.py tests/bash/test-controller-behavior-multi-host.sh docs-private/reviews/<date>-chunk-16/<host>-<scenario>.transcript.log

Usage

# Maintainer / live suite (not required for default ./tests/run.sh)
export GOALFLIGHT_CONTROLLER_BEHAVIOR=1

python3 scripts/hosts/controller/probe_matrix.py --json
python3 scripts/hosts/controller/behavior_scenario.py \
  --controller codex \
  --scenario doctor-loads \
  --directory "$(git rev-parse --show-toplevel)" \
  --json

python3 scripts/hosts/controller/behavior_scenario.py \
  --controller claude-acp \
  --scenario doctor-loads \
  --directory "$(git rev-parse --show-toplevel)" \
  --transcript-dir "docs-private/reviews/$(date +%F)-chunk-15" \
  --json

# Full live matrix. Empty env gate skips exit 0.
export GOALFLIGHT_LIVE_CONTROLLERS=codex,claude-acp
bash tests/bash/test-controller-behavior-multi-host.sh

Read the full file on GitHub · 143 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. yesterday First seen · 143 lines · 14 tokens per session scan A 30df848a6daa

Subscribe to this mod's changes

controller-behavior-test is a command published in the GitHub repository simonrowland/goal-flight (20 stars, last pushed 3d ago), licensed MIT. It adds 14 tokens to every session and 1,781 once invoked, about $0.0001 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.