SkillsBench challenge: skills off, decided; the final ranking uses SkillsBench

SkillsBench challenge; Terminal-Bench 2 out of the arena; gates check SkillsBench; practice off the board

Front page: no hero; the Space opens on the board

AGENTS.md: a Harbor dataset over 200 tasks is submitted as a subset

Merge branch 'arena/harbor-import' into arena/merge-0930

Board: open on Benchmarks, one benchmark's hill climb and leaderboard at a time

Benchmarks: rank and plot one benchmark at a time, per domain where it has domains

AGENTS.md: Harbor tasks are an accepted format

Harbor datasets validate and run as they are

Supported by OpenEnv: the mark uses the Hugging Face logo

Add your agent: Try it first with a pinned example; board posts only when asked; runs on the challenge's compute

Tests: every board sign-in entry point opens a new tab when framed

Front page opens with posttrain.com's hero, then the line plot, the leaderboard and the board

Static gates: S-VERIFIER-NETWORK also sees verify-time installs, sourced helpers, Python shell-outs, datasets and cloud copies (F12-02)

TMax page: PostTrain 0.1.11, its evals leave the leaky tasks out, costs from 0.1.11's dry runs (F12-01)

Add your agent: a clone address's .git comes off the dataset's name (F12-09)

Static gates: the grading-data message agrees in number when several answer-like files hold what the verifier checks

Tests run offline: HF Jobs stubbed in conftest.py, the mock world's stored files from fixtures

Static gates v3: an answer-like file that holds what the verifier checks is grading data

Console 0.1.11 (PostTrain main d2587fa): round-11 fixes

TMax page: "The guide's section Train on them and compare", as the link reads (F11-15)

TMax page: a first eval takes about 5 to 12 minutes; PostTrain 0.1.10; the suite before the training folder; the data extras; the guide's section linked (F11-11, F11-15)

arena_cli.py: a command that needs your identity and finds no token says so and exits 4, without the usage banner (F11-12)

Arena: challenges and the two lists have their own addresses; an unknown submission answers 404 (F11-10)

Starter kit: upload to the dataset the modal's token covers, no token in a command, the static gates in the local check (F11-08)

Add your agent: the dataset field takes the forms people paste and shows the id it will use (F11-06)

Static gates: S-VERIFIER-NETWORK also reports a verifier that fetches data offline (F11-03)

Console 0.1.10 (PostTrain main 8ab7d4d): round-10 fixes

Starter kit: the upload step installs hf the one way the rest of the Space does (uv tool install hf) and logs in before hf upload (F10-12)

TMax page: say which tasks an attempt can pass without solving them (F10-03, the page's part)

TMax page's PostTrain steps: a first eval of 4 tasks, the held-out tasks as a suite, cost and time from the dry runs, an indicative comparison (F10-05, F10-06, F10-07)

validate: a dataset or repository that isn't there says so, not the generic 422 (F10-08)

Add your agent: the prompt's address keeps its break points when the page fills in its own (a phone broke AG|ENTS.md)

A submission's own address, /arena/submissions/<id>, previews as the submission (F10-17)

Add your agent: the dataset first and a token that writes only to it; one way to get hf; the prompt wraps between words and a rewritten name shows its id (F10-12..15, F10-02, F10-09)

Static gates: a verifier that downloads while the task turns the network off (F10-02), and answer files the build writes (TMax task_000519)

/arena: a challenge's name wraps after a separator dot, not before; the starter kit's blocks wrap as before (scrolled, a phone showed cut words); the simulated-data banner's button at its right edge

/arena stylesheet: drop two unused tokens and rules a more specific one already covers

/arena: loading, a page that could not load and an empty table show the board's state panel (centered, mono, muted)

/dashboard/previous: its current page and tabs as the board's; 13 px text

/arena: what the fresh-eyes comparison with the board found

Board: its phone rules apply (they sat above the rules they override)

/arena: an outlined button keeps the board's outlined hover (the filled buttons' hover turned a hovered Check navy with dark text)

/dashboard/previous in the board's palette, faces and buttons

/arena in the board's style: its header, faces, palette and components

Supported by OpenEnv: a small linked mark on the board's header and the app's footer, and one sentence in the README and AGENTS.md

Board: Add your agent in four steps (a token, logging the agent in, an agent id, a prompt that gets the agent working); sign-in from inside huggingface.co opens the Space in a new tab

Console 0.1.9 (PostTrain main 47a3e88): round-9 fixes, task packages by what a run needs, one exact_match

app.py: the keep-alive comment says what was measured (it didn't change the rate of Hugging Face's 502 pages)

README: the short description and the first paragraph name the Space's two parts (F9-15).