antithesis-setup（对立关系设置）

Antithesis Setup

Purpose and Goal

Scaffold the antithesis/ harness needed to bring the system up in Antithesis in a mostly idle, ready state.

Success means:

• antithesis/config/docker-compose.yaml exists and required SUT images are referenced with build: directives
• snouty validate on antithesis/config/ succeeds
• the SUT dependency graph includes the relevant Antithesis SDK where assertions or lifecycle hooks will run
• at least one minimal bootstrap property exists in a simple SUT path and is expected to show up in the first Antithesis run
• The harness is ready for the antithesis-workload skill to add or iterate on test templates, assertions, and workload code
• If the user asks to submit or launch a run, use the antithesis-launch skill — do not run snouty run directly

Prerequisites

• Research artifacts required. Before proceeding, check whether antithesis/scratchbook/ exists and contains research output (at minimum sut-analysis.md and deployment-topology.md). If the scratchbook is missing or empty, stop and warn the user: The antithesis-research skill has not been run yet (no scratchbook found at antithesis/scratchbook/). Setup depends on research artifacts — especially the SUT analysis and deployment topology — to make informed decisions about instrumentation, image structure, and service composition. Please run antithesis-research first, review its output, then return to setup. Do not attempt to proceed without research artifacts. The setup skill will make significantly worse decisions without the context that research provides.
• DO NOT PROCEED if snouty is not installed. See https://raw.githubusercontent.com/antithesishq/snouty/refs/heads/main/README.md for installation options.

Documentation Grounding

Use the antithesis-documentation skill to access these pages. Prefer snouty docs.

• Docker Compose setup guide: https://antithesis.com/docs/getting_started/setup/
• Docker best practices: https://antithesis.com/docs/best_practices/docker_best_practices/
• Coverage instrumentation: https://antithesis.com/docs/instrumentation/coverage_instrumentation/
• Assertion cataloging: https://antithesis.com/docs/instrumentation/assertion_cataloging/
• Handling external dependencies: https://antithesis.com/docs/reference/dependencies/
• Fault injection: https://antithesis.com/docs/environment/fault_injection/

Workflow

This skill is broken out into multiple steps, each in a different reference file. Read and implement each reference file listed below one at a time to fully set up a project. After implementing each step, check whether what you learned invalidates any decisions from earlier steps. Instrumentation decisions (step 2) are the most common thing that needs revision once you start building images (step 3).

• references/directory-init.md: initialize or merge the antithesis/ directory from assets/antithesis/
• references/instrumentation.md: decide how each SUT service is instrumented, how the SDK is installed, how symbols are delivered, and where the bootstrap property lives
• references/docker-images.md: create or adapt Dockerfiles for SUT components
• references/docker-compose.md: write antithesis/config/docker-compose.yaml
• references/config-dir.md: understand what belongs in antithesis/config/
• references/submit-and-test.md: test locally and submit the first run

General Guidance

• Merge with existing antithesis/ content instead of overwriting it.
• Prefer podman compose for local testing; fall back to docker compose.
• Keep Antithesis-only scaffolding under antithesis/ when practical.
• Focus this skill on infrastructure and readiness, not on defining the workload itself.
• Installing the relevant Antithesis SDK into the SUT and adding one minimal bootstrap assertion is part of setup, not deferred workload work.
• If antithesis/test/ does not exist yet, create the directory structure needed for later workload work, but leave real test templates and assertions to antithesis-workload.
• Treat instrumentation and symbolization as bootstrap work. The setup is not complete until the relevant images expose /opt/antithesis/catalog/ or /symbols/ correctly for their language.
• Treat local testing as required before the first submission.
• Use snouty run directly to submit runs. Run compose build before snouty run to ensure images are up to date.
• Do not add a separate Dockerfile under antithesis/config/ unless the deployment explicitly requires it.
• Disable color/ANSI output in every container. Antithesis stores raw bytes and does not render escape codes — color output is garbage in logs and triage. Set NO_COLOR=1 on all services via docker-compose.yaml environment blocks or Dockerfile ENV directives. Add tool-specific flags (e.g. FORCE_COLOR=0) where needed.

Self-Review

Before declaring this skill complete, review your work against the criteria below. If your agent supports spawning sub-agents, create a new agent with fresh context to perform this review — give it the path to this skill file and have it read all output artifacts. A fresh-context reviewer catches blind spots that in-context review misses. If your agent does not support sub-agents, perform the review yourself: re-read the success criteria at the top of this file, then systematically check each item below against your actual output.

Review criteria:

• antithesis/config/docker-compose.yaml exists and every service has build: (for local images) or image: (for public images) configured correctly
• Every service in docker-compose.yaml includes platform: linux/amd64
• Every service has hostname: set to match its container_name:
• The instrumentation inventory from references/instrumentation.md is fully implemented: each service is instrumented, cataloged-only, or explicitly documented as uninstrumented
• The relevant Antithesis SDK is installed in the SUT dependency graph
• A bootstrap property exists in a simple, guaranteed-to-run code path (not behind rare behavior)
• /opt/antithesis/catalog/ or /symbols/ is exposed correctly for each service's language
• The setup_complete signal is wired in at least one entrypoint
• snouty validate on antithesis/config/ succeeds
• All built images target amd64 (verified via podman image inspect or docker image inspect)
• Every service has NO_COLOR=1 set in its environment (via docker-compose.yaml and/or Dockerfile) to prevent ANSI escape codes in container output
• The harness is ready for the antithesis-workload skill — test template directories exist or are wired for later use
• If the user asks to launch a run, the antithesis-launch skill is used instead of running snouty run directly