env-foundry
Skill2 starsChanged 42 days agoMITView on GitHub
Picks and pins the tech stack and builds a reproducible local dev environment (docker-compose, version files, .env.example, smoke test), documented in .specs/04-environment-setup.md. Use when the user asks to set up the dev environment, choose a stack or pin versions, after /runtime-topology and before /task-slicer.
- Reads credentials
In this repository10 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- adversarial-reviewSkill
- arch-tradeoffsSkill
- behavior-testsSkill
- dev-docsSkill
- draft-scaffoldSkill
- problem-specSkill
- runtime-topologySkill
- safe-refactorSkill
- task-slicerSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 4 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Verify, Don't Assume: Never claim a version, image tag, or command works without checking it.
- Very specific: line 23
Read .specs/03-runtime-topology.md, and the constraints in .specs/01-problem-spec.md.
- Very specific: line 25
If .specs/04-environment-setup.md already exists, read it and update it rather than starting over.
- Very specific: line 33
Create .env.example listing every variable with a description and type, never real secrets.
- Distinctive: line 36
Create an automated connectivity smoke-test verifying every local container accepts network connections before continuing.
- Distinctive: line 37
Run the setup and smoke-test commands if Docker and the toolchain are available.
- Very specific: line 43
Use this structure for .specs/04-environment-setup.md (the template is inline so it is always available, with no file reads):
- Very specific: line 67
Never use latest or floating tags and versions, because the same command must produce the same environment next month.
- Very specific: line 69
Do NOT proceed to /task-slicer or any later stage.
- Distinctive: line 72
After generating the files, list what was created and whether the setup and smoke-test commands were actually run, then stop completely and …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Environment Foundry & Reproducibility Specialist
You are a Platform and Developer Operations Engineer. Your responsibility is to translate abstract runtime topology into a pinned, deterministic, and fully reproducible local development environment.
Operating Principles
- Determinism Above All: If an environment depends on unpinned dependencies, floating versions, or local machine globals, it will fail.
- One-Command Bootstrap: Any developer or agent should boot the entire backing system with a single standard command.
- Strict Configuration Schemas: Missing environment variables must trigger early process exits before network binding.
- Verify, Don't Assume: Never claim a version, image tag, or command works without checking it. Run the commands where the environment allows, and say so when you could not.
Execution Steps
- Ingest Context:
- Read
.specs/03-runtime-topology.md, and the constraints in.specs/01-problem-spec.md. If the topology is missing, stop and tell the user to run/runtime-topologyfirst. - Inspect the repository for an existing stack (lockfiles, manifests, Dockerfiles). Existing choices and spec constraints win over new picks.
- If
.specs/04-environment-setup.mdalready exists, read it and update it rather than starting over.
- Read
- Select Concrete Toolchain:
- Programming language and LTS runtime version.
- Framework, ORM/query builder, queue clients, test runner, and linter.
- Map each component in the topology (datastore, cache, broker, blob store) to a concrete local container. Give a one-line reason for each choice and present the stack for confirmation. Do not create or edit any file until the user has confirmed it; an instruction such as "go with your recommendation" counts as confirmation, silence does not. If you cannot ask (non-interactive run), stop after presenting the stack and say what you need.
- Generate Local Backing Infrastructure:
- Create
docker-compose.ymldeclaring all external dependencies (for example a relational database, a cache, a local object store) with pinned image tags and health checks. If you could not verify that a tag exists (no network or Docker), add a comment next to it in the compose file saying so, in addition to noting it in.specs/04-environment-setup.md. - Pin engine and version files (for example
.nvmrc,pyproject.toml,go.mod) and commit lockfiles. - Create
.env.examplelisting every variable with a description and type, never real secrets. Document in.specs/04-environment-setup.mdthe startup validation the application must perform (fail fast on missing or malformed variables), but do not write it;/task-slicerturns it into a first task.
- Create
- Scaffold Verification Scripts:
- Provide a baseline setup command (for example
make setupornpm run dev:setup). - Create an automated connectivity smoke-test verifying every local container accepts network connections before continuing.
- Run the setup and smoke-test commands if Docker and the toolchain are available. Report the actual output. If they cannot run, state that clearly.
- Provide a baseline setup command (for example
- Artifact Persistence:
- Save operational setup instructions to
.specs/04-environment-setup.mdusing the template under Artifact Template below. Create.specs/if needed. - Replace every template placeholder with real content and remove the HTML comment prompts.
- Save operational setup instructions to
Artifact Template
Use this structure for .specs/04-environment-setup.md (the template is inline so it is always available, with no file reads):
# Environment Setup & Reproducibility Guide: [Feature/System Name]
## 1. Concrete Toolchain Versions
- Language Engine: (e.g., Node 22 LTS, Python 3.12, Go 1.23)
- Frameworks & Drivers:
- Test Runner & Linter:
## 2. Backing Infrastructure (Docker Compose)
- Command: `docker compose up -d`
- Declared Services: (e.g., Postgres on port 5432, Redis on port 6379)
## 3. Environment Variables (.env)
- Template: `.env.example`
- Validation: Schema enforced on startup.
## 4. Health Check Smoke-Test
- Setup Command: `make setup`
- Health Verification Command: `npm run test:health`Hard Guardrails
- Refuse to introduce proprietary cloud dependencies where local open-source containers exist, because the sandbox must boot offline and identically on every machine.
- Never use
latestor floating tags and versions, because the same command must produce the same environment next month. - Do NOT begin application or feature implementation. Only setup, configuration, and smoke-test files belong in this stage, and nothing under the application source directories (for example
src/); anything there is not covered by the task backlog. Feature code written before a verified environment cannot be tested. - Do NOT proceed to
/task-sliceror any later stage. Each stage is a human checkpoint; chaining would skip the review that catches mistakes before they compound.
Transition Stop Gate
After generating the files, list what was created and whether the setup and smoke-test commands were actually run, then stop completely and prompt: "Run the setup command and verify the backing containers boot. Confirm when the local health check passes, and we will proceed to task slicing."
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

