feat: vf v1 <> nano bridge

[mikasenghaas]

• README - high level overview
• GUIDE - user guide to authoring taskset + harness, cli usage, etc.
• ARCHITECTURE - explanation of framework internals

[macroscopeapp]

🟢 Low v1/GUIDE.md:280

The bash examples on lines 280-285 place inline comments (# per-rollout budgets, etc.) after \ line continuations. In bash, \ must be immediately followed by a newline to continue the line — any trailing space or comment causes a parse error when the command is copy-pasted. Consider moving the comments above each line or removing them from the continuation lines.

 ```bash
 uv run eval gsm8k-v1 -n 5 -r 3 \
-  --max-turns 8 --max-total-tokens 8192 \                          # per-rollout budgets
-  --retries.model.max-retries 3 --retries.runtime.max-retries 3 \  # retry one failed call
-  --retries.rollout.max-retries 3 --retries.rollout.include ProgramError \  # retry a whole rollout, by error type
-  --timeout.rollout 600 --timeout.scoring 120                      # per-stage wall-clock caps (seconds)
+  --max-turns 8 --max-total-tokens 8192 \
+  --retries.model.max-retries 3 --retries.runtime.max-retries 3 \
+  --retries.rollout.max-retries 3 --retries.rollout.include ProgramError \
+  --timeout.rollout 600 --timeout.scoring 120

<details>
<summary>🚀 Reply "<strong>fix it for me</strong>" or copy this <strong>AI Prompt</strong> for your agent:</summary>

```text
In file @verifiers/v1/GUIDE.md around lines 280-285:

The bash examples on lines 280-285 place inline comments (`# per-rollout budgets`, etc.) after `\` line continuations. In bash, `\` must be immediately followed by a newline to continue the line — any trailing space or comment causes a parse error when the command is copy-pasted. Consider moving the comments above each line or removing them from the continuation lines.

[macroscopeapp]

🟢 Low servers/wiki.py:42

tarfile.extractall(filter="data") raises TypeError on Python versions before 3.11.4 (and before 3.10.12), because the filter keyword argument did not exist yet. This crashes setup at runtime on those interpreters. Consider wrapping with a try/except TypeError fallback.

-                    t.extractall(cache, filter="data")
+                    try:
+                        t.extractall(cache, filter="data")
+                    except TypeError:
+                        t.extractall(cache)

🚀 Reply "fix it for me" or copy this AI Prompt for your agent:

In file @environments/wikispeedia_v1/wikispeedia_v1/servers/wiki.py around line 42:

`tarfile.extractall(filter="data")` raises `TypeError` on Python versions before 3.11.4 (and before 3.10.12), because the `filter` keyword argument did not exist yet. This crashes `setup` at runtime on those interpreters. Consider wrapping with a `try`/`except TypeError` fallback.

Evidence trail:
- environments/wikispeedia_v1/wikispeedia_v1/servers/wiki.py line 42: `t.extractall(cache, filter="data")` at REVIEWED_COMMIT
- pyproject.toml line 14: `requires-python = ">=3.11,<3.14"` at REVIEWED_COMMIT
- Python 3.11 docs (https://docs.python.org/uk/3.11/library/tarfile.html): 'Нове в версії 3.11.4' (New in version 3.11.4) for extraction filters
- CPython issue #102950 tracking backport to 3.11: https://github.com/python/cpython/issues/102950
- Python docs recommend `hasattr(tarfile, 'data_filter')` for compatibility checking

[macroscopeapp]

🟢 Low cli/init.py:46

When name is whitespace-only (e.g., " ") or only slashes (e.g., "/"), the strip() calls on line 50 return an empty string, so pkg becomes empty. This causes env_dir = Path(config.path) / pkg to resolve to the parent directory itself (e.g., ./environments), and scaffolded files like __init__.py and taskset.py are written there instead of a package subdirectory. Consider validating that pkg is non-empty after processing and raising an error for invalid names.

def _names(name: str) -> tuple[str, str, str, str]:
     """`(dash, pkg, stem, prefix)` derived from a raw name: the hyphenated id, the importable
     package (underscores), the `_v1`-less stem (for tool prefixes), and the CamelCase class
     prefix (e.g. `my-task-v1` -> `my-task-v1`, `my_task_v1`, `my_task`, `MyTask`)."""
     dash = name.strip().strip("/").replace("_", "-").lower()
+    if not dash:
+        raise ValueError(f"invalid environment name: {name!r}")
     pkg = dash.replace("-", "_")
     stem = pkg[:-3] if pkg.endswith("_v1") else pkg
     prefix = "".join(part[:1].upper() + part[1:] for part in stem.split("_") if part)
     if not prefix or not prefix[0].isalpha():
         prefix = f"Env{prefix}"
     return dash, pkg, stem, prefix

🚀 Reply "fix it for me" or copy this AI Prompt for your agent:

In file @verifiers/v1/cli/init.py around lines 46-56:

When `name` is whitespace-only (e.g., `"   "`) or only slashes (e.g., `"/"`), the `strip()` calls on line 50 return an empty string, so `pkg` becomes empty. This causes `env_dir = Path(config.path) / pkg` to resolve to the parent directory itself (e.g., `./environments`), and scaffolded files like `__init__.py` and `taskset.py` are written there instead of a package subdirectory. Consider validating that `pkg` is non-empty after processing and raising an error for invalid names.

Evidence trail:
verifiers/v1/cli/init.py lines 46-56 (_names function), line 289 (env_dir = Path(config.path) / pkg), line 290 (pkg_dir = env_dir / pkg), line 339 (if not config.name: check on raw name, not processed pkg). Python Path behavior: Path('a') / '' resolves to Path('a').