ProjectContext API

The project object passed to every standard's evaluate function provides these methods:

Methods

`project.get_file_content(path) -> str | None`

Read the content of a file by its path. Returns None if the file doesn't exist.

content = project.get_file_content("README.md")
if content and "## Contributing" in content:
    # Has a contributing section

`project.get_file_size(path) -> int | None`

Size of a file in bytes, including files you cannot read as text (images, audio, wasm). Known for artifact projects; None on Git repositories.

too_big = [f for f in project.list_files() if (project.get_file_size(f) or 0) > 5_000_000]

`project.read_file_lines(path, start=1, count=100) -> dict | None`

Part of a file as numbered lines, for reading large or minified files a page at a time. Lines longer than 200 characters (a minified bundle is often one line) are first broken after ;, ,, braces or parentheses. Each call returns about 5,000 characters at most and says where to continue (next_start, None at the end). This is how an LLM standard's model reads big files; get_file_content stays the raw file.

page = project.read_file_lines("assets/app.js")
# {"path": "assets/app.js", "total_lines": 2140,
#  "lines": ["1: (function(){\"use strict\";", ...], "next_start": 61}

`project.list_files() -> list[str]`

List all file paths in the repository.

files = project.list_files()
has_ci = any(f.startswith(".github/workflows/") for f in files)

`project.get_languages() -> dict[str, float]`

Language breakdown as percentages.

languages = project.get_languages()
# {"Python": 72.5, "Dockerfile": 15.0, "YAML": 12.5}

`project.get_members() -> list[dict]`

Project members with roles.

members = project.get_members()
# [{"username": "alice", "role": "maintainer"}, ...]

`project.get_contributors() -> list[dict]`

Contributors with commit counts.

contributors = project.get_contributors()
# [{"username": "alice", "commits": 142}, ...]

`project.get_default_branch() -> str`

Name of the default branch.

branch = project.get_default_branch()  # "main"

`project.get_topics() -> list[str]`

Repository topics/tags.

topics = project.get_topics()
# ["python", "backend", "api"]

`project.get_metadata() -> dict`

General repository metadata.

meta = project.get_metadata()
# {
#   "name": "my-project",
#   "description": "A cool project",
#   "web_url": "https://github.com/org/my-project",
#   "created_at": "2024-01-01T00:00:00Z",
#   "updated_at": "2024-06-01T00:00:00Z",
# }

`project.get_file_last_commit_date(path: str) -> str | None`

ISO 8601 timestamp of the most recent commit that touched path on the default branch. Returns None if the file does not exist or has no commit history. Useful for staleness / freshness checks on specific tracked files.

date_str = project.get_file_last_commit_date("CLAUDE.md")
# "2026-04-15T10:23:00Z"

Execution environment

Standards run in isolated gVisor containers with no network access
Execution timeout: 30 seconds
Memory limit: 128 MB
Only the project API is available — no filesystem, no imports beyond the standard library
For an artifact project, project reads the uploaded commit's files; members, contributors, topics and commit dates are empty