star-env-builder

name	star-env-builder
disable-model-invocation	true
description	Build and verify the project's Python runtime environment so plan execution has a working interpreter. Reads .env: a valid CONDA_HOME creates conda env ENV_NAME (argument, default CODE_NAME); otherwise a .venv under the project root. An existing environment is never deleted — after user confirmation it is renamed to a dated backup (real run date) before rebuilding. Dependencies resolve first-signal-wins: existing CODE_NAME/requirements* → packaging metadata (pyproject / setup.py / environment.yml) → import scan of the code, with generated results written to a two-tier layout (requirements.txt referencing requirements/framework|runtime|optional.txt; conda-only items in requirements/conda.txt). Installs through a uv > pip > conda ladder with CUDA-aware framework wheel selection behind a single install-plan gate, then smoke-tests in three layers (imports → framework/GPU → project entrypoint) and writes ENV_REPORT.md plus a version freeze under wkdrs/. Use when the user runs /star-env-builder, wants the project's conda env or venv created or rebuilt, needs dependencies resolved and installed, or wants the runtime environment verified. Bilingual (en/zh).

Research Env Builder — runtime environment bootstrap

Match the user's language. For Chinese dialogue, read SKILL_zh.md in full before acting and follow it as the localized instructions; load other *_zh.md resources when referenced. Otherwise, follow this file and load unsuffixed resources. If SKILL_zh.md conflicts with this file, this SKILL.md is authoritative.

Invocation: /star-env-builder [ENV_NAME | add <package>…] — the conda environment name to create, omitted to use CODE_NAME from .env; add installs one or more packages into the environment .env already names and records them in the requirements layout.

Shared conventions. Read docs/mds/star-workflow/research-workflow-conventions.md (Chinese: research-workflow-conventions.zh-CN.md) before acting: §1 git, §2 the STOP line, §3 .env runtime, §4 real dates, §5 plan-name resolution, §6 delegation, §7 dialogue, §8 the artifact registry, §9 project layout. It is the baseline every STAR skill shares; this file states what is specific to this one, and wins wherever it is stricter.

Role

You give the codebase a working runtime. Upstream, star-code-architect lands ${CODE_NAME}/ but stops at the environment — its runtime-smoke step prepares install commands and hands them to the user (STOP line). Downstream, star-plan-executor runs every command through the .env environment and assumes it works. This skill produces that environment: a conda env or .venv resolved from .env, a dependency layout under ${CODE_NAME}/requirements/ when one was missing, and an evidence-backed environment report under wkdrs/.

You build the environment; you do not implement or refactor research code. The only writes into ${CODE_NAME}/ are generated requirements files. If code changes are needed to make the project importable, hand off to star-plan-executor.

Core Principles

1. .env is the only path source; never activate (conventions §3). Resolve the target interpreter once — ENV_PY = $CONDA_HOME/envs/<ENV_NAME>/bin/python or <project>/.venv/bin/python — and run everything through that absolute path. This skill owns the environment: it is the only one that may create, rename, or install into one.
2. One gate; situational asks. The single gate is install-plan approval (Step 4): nothing installs before it; everything it covers runs autonomously after it. Situational questions — overwrite an existing env, a CUDA mismatch, uv missing, a conda-only dependency under a venv backend — are asked when hit, via AskUserQuestion, one question per call, each with a recommendation.
3. Rename, never delete. An existing environment is backed up by renaming to <name>_<YYYYMMDD> — the date from date +%Y%m%d at run time, never invented. This skill deletes no environment, ever; stale backups are the user's to clean.
4. Category is policy; the ladder is uv > pip > conda. framework (CUDA-coupled, index-pinned) / runtime (ordinary PyPI) / optional (logging, viz, dev extras) / conda.txt (system-isolation items). Each category has its own install route and failure handling: prefer uv, fall back to pip per package, use conda only for the whitelist and only under a conda backend. Policy: references/installer_policy.md.
5. Adopt what exists; generate only what is missing. An existing requirements layout is installed as-is, never rewritten. Generated dependencies come from packaging metadata before import scanning (references/dependency_resolution.md), land in the two-tier layout, and are committed as a code asset once the build is verified.
6. Evidence-based acceptance. The main loop runs the three smoke layers itself (references/smoke_test_spec.md) and reports what was verified with evidence, not that it "works" (CLAUDE.md §7). Chats end, files do not: the report and version freeze go to wkdrs/env_<ENV_NAME>_<date>/.

Workflow

Step 0: Preflight

1. Read .env and resolve CODE_NAME, CONDA_HOME, PYTHON_HOME (conventions §3).
2. ENV_NAME := the argument, else CODE_NAME. An add <package>… argument instead selects add mode: skip to Step 8, targeting the environment .env already names — nothing is created, renamed, or rebuilt.
3. Probe and record (this feeds the install plan and the report): platform + arch; nvidia-smi (driver's CUDA ceiling); nvcc --version / CUDA_HOME (local toolkit, often absent); $CONDA_HOME/bin/conda --version; uv --version.
4. ${CODE_NAME}/ missing or effectively empty → there is no dependency source; recommend /star-code-architect first, and offer to build a bare env (python only) if the user wants one anyway.

Step 1: Choose the backend (deterministic)

• CONDA_HOME non-empty and the path exists → conda backend: $CONDA_HOME/bin/conda create -n <ENV_NAME> python=<X.Y> -y.
• Otherwise → venv backend at <project>/.venv: prefer uv venv .venv --python <X.Y>; else $PYTHON_HOME/bin/python -m venv .venv; last resort python3 -m venv .venv. An ENV_NAME argument is meaningless here — say so if one was passed, then continue.
• Python version: requires-python (pyproject.toml) → python_requires (setup.py / setup.cfg) → the upstream README's stated version → default 3.10. Conflicting signals → ask.
• Record ENV_PY (absolute path) and use it for every later command.

Step 2: Collision handling

• conda: <ENV_NAME> already in conda env list → ask one question with three options: backup & rebuild (rename to <ENV_NAME>_$(date +%Y%m%d) via conda rename; older conda lacking rename: create --clone + remove, warn that disk usage doubles temporarily) / verify & repair in place (skip creation; jump to Step 5 for failed items or Step 6 — the resume path when a previous run was interrupted) / abort (exit cleanly, nothing touched).
• venv: .venv exists → same three-way ask → backup is mv .venv .venv_$(date +%Y%m%d). Note in the report: a moved venv has its old absolute paths baked into scripts — it is a frozen backup for reference and rollback, not an activatable environment.
• Backup name already taken → append -<HHMM> (also from date).

Step 3: Resolve dependencies (first signal wins)

Recipe and mapping table: references/dependency_resolution.md.

1. ${CODE_NAME}/requirements.txt or ${CODE_NAME}/requirements/ exists → adopt as-is; never rewrite, reorder, or "improve" it.
2. Else packaging metadata — pyproject.toml [project.dependencies], setup.py / setup.cfg install_requires, environment.yml — transcribe into the two-tier layout, keeping every version constraint.
3. Else import scan: AST top-level imports over ${CODE_NAME}/ → drop stdlib and local modules → map import names to PyPI distributions (verify unknowns on PyPI) → write the layout, versions unpinned except known-coupled pairs.

Generated layout: requirements.txt holds only -r requirements/framework.txt and -r requirements/runtime.txt lines (optional referenced as a comment); requirements/framework.txt opens with the matched --extra-index-url; conda-only items go to requirements/conda.txt with a "conda installs this, not pip" header. Files are written now, committed in Step 7 after the build is verified.

Step 4: Gate — the user approves the install plan

Present as normal text: backend + env name + python version; dependency source used; per-category package counts and notable pins; the torch↔CUDA match (detected driver ceiling vs chosen wheel index); rough download size of the big wheels; conda.txt items; anything already flagged uncertain (CUDA mismatch, unresolved imports, version conflicts). Then ask via AskUserQuestion: approve and build / adjust (say what) / abort. Uncertainties are settled here — never silently.

Step 5: Install (tiered ladder)

Policy, whitelist, and wheel-index matrix: references/installer_policy.md. Order: conda.txt (conda backend only) → framework.txt → runtime.txt → optional.txt (only if the approved plan included it) → editable project install (--no-deps -e) when packaging metadata exists.

• uv present → uv pip install --python $ENV_PY -r <file>; uv absent → ask once: install uv / use pip for this run.
• Per-package failure → retry via pip (≤2 attempts total per package) → still failing: record it, continue with the rest, resolve or hand over at the end.
• venv backend hits a conda-only item → stop and ask: user installs it system-wide themselves / skip it / use a pip alternative if one exists.
• Source-build items (flash-attn and friends) → STOP line: prepare the exact command in the report; do not run it.
• Respect PIP_INDEX_URL / UV_DEFAULT_INDEX already configured; never override the user's mirrors, never write global config.

Step 6: Smoke test (three layers, run by the main loop)

Spec and evidence format: references/smoke_test_spec.md.

• L1 imports: every distribution in framework + runtime (and installed optional) imports and reports a version through $ENV_PY.
• L2 framework: torch.cuda.is_available() + device count + a small tensor op on the device (mps on macOS; CPU-only boxes noted as expected, not failed).
• L3 project: $ENV_PY -m compileall -q ${CODE_NAME}; then import <package> if editable-installed, else the cheapest entrypoint (--help, or pytest --collect-only -q). No data, no weights, no downloads — minutes, not hours.

A failed layer → diagnose from the traceback, fix (a missing transitive dep goes into the right generated requirements file), re-run the layer; ≤2 fix rounds per layer → still failing: mark it blocked with the error tail and continue where independent.

Step 7: Report, snapshot, commit

1. Write wkdrs/env_<ENV_NAME>_<YYYYMMDD>/ENV_REPORT.md from assets/env_report_template.md: identity + ENV_PY, machine probe, backup renames, per-category install results, the smoke matrix with evidence, failures/blocked items, awaiting-user commands.
2. uv pip freeze --python $ENV_PY (or $ENV_PY -m pip freeze) → freeze.txt alongside the report.
3. Requirements files generated this run (including deps added during smoke diagnosis) are committed now: star-env-builder: add requirements layout, staging only ${CODE_NAME}/requirements*.
4. .env's PYTHON_HOME does not resolve to the just-verified ENV_PY → downstream skills resolve the runtime from .env: offer to point PYTHON_HOME at the environment just built (conda: $CONDA_HOME/envs/<ENV_NAME>; venv: <project>/.venv) — only with explicit confirmation.
5. Chat report ≤400 words: what was verified (with evidence), failures, awaiting-user commands. Hand off downstream: /star-plan-executor <leaf> now has a runtime; /star-flow-status shows what to run next.

Step 8: Add packages (add mode only)

The environment already exists; this mode installs into it and records what it installed. It creates, renames, and rebuilds nothing — a broken environment is a full run's job (Step 2's verify & repair in place).

1. Resolve ENV_PY from .env (Principle 1). No usable interpreter → say so and recommend a full /star-env-builder run; install nothing.
2. Categorise each package per references/installer_policy.md — framework / runtime / optional / conda-only — and say which requirements file each will land in.
3. Gate (Principle 2 — nothing installs before it): present the packages, their categories, the versions and index that will be used, the download size when it is large, and any CUDA coupling; ask approve and install / adjust / abort.
4. Install through the ladder (uv > pip > conda; conda only under a conda backend and only for the whitelist). A source-build item stays on the STOP line: prepare the exact command, do not run it.
5. Smoke the new packages only (references/smoke_test_spec.md): L1 — each imports and reports a version through $ENV_PY; a new framework package also gets L2. A failure → diagnose, one bounded retry, then mark it blocked and report; never leave a package installed but unverified.
6. Append each installed package to its requirements file, preserving the layout's existing order and pins. Append an ## Added <date> block to the newest wkdrs/env_<ENV_NAME>_<date>/ENV_REPORT.md (none exists → write a fresh report). Commit: star-env-builder: add <packages>, staging only ${CODE_NAME}/requirements*.
7. Report ≤400 words: what installed, what each requirements file gained, the smoke evidence, anything blocked or awaiting the user.

State & File Rules

• Writes are limited to: the environment itself (under $CONDA_HOME/envs/ or <project>/.venv), ${CODE_NAME}/requirements* (only when generating a missing layout or filling a verified gap), wkdrs/env_<ENV_NAME>_<date>/, and — only with explicit user confirmation — the PYTHON_HOME= line in .env. Never touch source code, metds/plans/*, or other skills' outputs.
• Never delete an environment; backups are renames stamped with the real run date. Never invent timestamps.
• Git: at most one commit per run — requirements generated, or packages added in add mode — staging only ${CODE_NAME}/requirements* (conventions §1).
• Gate-approved installs run autonomously, including framework-scale downloads. STOP line regardless of approval: sudo or system package managers (apt / brew), driver or CUDA-toolkit system installs, CUDA source compilation (flash-attn-style builds), downloads over ~10 GB, deleting any environment. Prepare those as exact commands in the report instead.
• Respect the user's mirror configuration (PIP_INDEX_URL, UV_DEFAULT_INDEX); never write pip config, .condarc, or uv.toml.
• Re-invoke semantics: if a matching wkdrs/env_<ENV_NAME>_*/ENV_REPORT.md exists and the env is present, prefer verify & repair in place (Step 2) — resume from its failures instead of rebuilding.

Dialogue Discipline

• The gate and all situational questions go through AskUserQuestion — one question per call, each with a recommendation. If it is unavailable (headless / scripted), fall back to plain text, still one at a time; the install plan then needs an explicit approval message before anything installs.
• ENV_REPORT.md body language follows the dialogue language; keep technical terms in English inside Chinese reports.