Skip to content

rainboyOJ/problems / .agents/skills/roj

ROJ Client

Use scripts/roj.py to interact with the read-only ROJ site. The website serves problem metadata, Markdown/PDF statements, and public test data. It does not accept accounts, source-code submissions, or remote judging requests.

Run the globally installed script from any directory with Python 3:

python3 ~/.agents/skills/roj/scripts/roj.py --help

Set ROJ_URL or pass --base-url when using a local or staging server:

ROJ_URL=https://roj.ac.cn python3 ~/.agents/skills/roj/scripts/roj.py list
python3 ~/.agents/skills/roj/scripts/roj.py --base-url http://127.0.0.1:3033 list

Commands

Find problems

list searches by numeric ID or title. It automatically follows all result pages; use --page N when only one page is needed.

python3 ~/.agents/skills/roj/scripts/roj.py list
python3 ~/.agents/skills/roj/scripts/roj.py list "最短路"
python3 ~/.agents/skills/roj/scripts/roj.py list --page 2
python3 ~/.agents/skills/roj/scripts/roj.py --json list "二分"

Use --json for tool or AI consumption. JSON output is also available after a subcommand, for example list --json.

Read a problem

get prints metadata followed by the raw Markdown statement. PDF-only problems print the browser URL instead; the CLI does not attempt to convert PDFs to text.

python3 ~/.agents/skills/roj/scripts/roj.py get 1000
python3 ~/.agents/skills/roj/scripts/roj.py get 10000 --no-markdown
python3 ~/.agents/skills/roj/scripts/roj.py --json get 1000

Download public data

ROJ publishes each problem's public data as a single packed archive, never as individually fetchable files: the manifest lists archive.name (e.g. 1337.tar.zst) plus a files index, and any per-file URL built from files[].path returns 404. download therefore fetches the archive from the first source that answers, keeps it beside the data directory, unpacks it, and verifies every manifest path landed.

The default destination is ./roj-data/<id>/data/. The archive itself is stored next to that directory (./roj-data/<id>/<id>.tar.zst), so re-running download reuses it instead of hitting the network again; pass --force to re-fetch. --output names a destination directory (the archive is stored inside it unless it is named data).

python3 ~/.agents/skills/roj/scripts/roj.py download 1000
python3 ~/.agents/skills/roj/scripts/roj.py download 1000 --output ./data/1000
python3 ~/.agents/skills/roj/scripts/roj.py download 1000 --force

Sources are tried in the manifest's sources order until one returns the archive. If the archive carries an extra top-level directory (commonly data/), it is stripped during extraction so files land exactly at destination/<path>; a missing manifest file after extraction is an error, not a silent success. Requires tar and zstd (tar --zstd first, zstd -dc | tar -x as fallback).

Test C++ locally

test is local-only. It compiles with g++ -std=c++17 -O2, runs each matching .in/.out or .in/.ans pair, applies the problem time limit when present, and compares output with diff -b. A non-zero program exit is RE; timeout is TLE; different output is WA; all matching cases are AC. Memory limits are reported but are not forcibly imposed by this Python client.

Data lookup is offline by default, in this order:

  1. --data-dir DIR
  2. ./roj-data/<id>/
  3. ./roj/<id>/data/

If no data exists, add --download to explicitly fetch it into ./roj-data/<id>/. This is the only test mode that uses the network.

python3 ~/.agents/skills/roj/scripts/roj.py test 1000 main.cpp
python3 ~/.agents/skills/roj/scripts/roj.py test 1000 main.cpp --data-dir ./tests/1000
python3 ~/.agents/skills/roj/scripts/roj.py test 1000 main.cpp --download
python3 ~/.agents/skills/roj/scripts/roj.py --json test 1000 main.cpp

The command returns exit status 0 only when compilation succeeds and every data pair is AC. A missing pair, compiler error, network failure, or test failure is non-zero. JSON test results include the problem metadata, selected data directory, compiler details, per-case status, elapsed time, and any diagnostics.

Safe operating rules

  • Treat IDs and paths as untrusted input. Never construct a shell command from a problem title or a manifest filename.
  • Download only the problem's public archive listed by /api/problem/<id>/data; the bytes come from its configured sources. Do not expose or copy std.cpp, solution files, generators, checkers, or legacy down/ files.
  • Do not claim that a local AC is an official submission result. It only means the program matched the public data with this client's comparison rule.
  • Keep downloaded data in a user-selected working directory and inspect it before running unfamiliar source code.

API reference

The client uses these read-only endpoints:

GET /api/problems?q=<query>&page=<number>
GET /api/problem/<id>
GET /api/problem/<id>/markdown
GET /api/problem/<id>/data

/api/problem/<id>/data returns archive, sources and a files array. archive holds the packed file name (<id>.tar.zst) — that archive is the only downloadable artifact; appending a files[].path to a source baseUrl returns 404. Each source contains an id, a display label, and a problem-specific baseUrl, and sources are tried in order as <baseUrl><archive.name>. The files array is an index of what the archive contains (path + size), used to verify extraction and to let test --download decide whether data is already present.

The problem metadata response contains id, title, time, memory, tags, source, statementKind, hasMarkdown, hasPdf, and hasPublicData. A missing Markdown statement on a PDF-only problem is expected, not an API failure.

slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 192kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 192kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 192kUpdated Claude CodeCodexVerifiedInstall