Team/hugo

[22f1000808]

Team : HUGO

Solo Member: Shivam Mishra

Framework

Go's net/http with http.ServeMux (Go 1.22 method-prefixed routes) — no framework dependencies, fewer moving parts to audit when the whole point is sandbox isolation.

How to run locally

Clone the repo, run git submodule update --init to fetch nsjail, then make docker-run to build and start the container. The Makefile includes make test for unit tests, make integration to run the e2e test suite, and make check for the full pre-submission verification. Smoke-test with curl localhost:8080/healthz.

Security

All documented attack surfaces are closed:

1. Path traversal: internal/validator/filename.go validates filenames as single components; internal/runner/runner.go enforces filepath.Clean at the boundary.
2. Shell command injection: No exec.Command("sh") anywhere. Workspace creation uses os.MkdirTemp and cleanup uses os.RemoveAll.
3. Compiler flag injection: internal/config/flags.go and internal/validator/flags.go enforce per-language allow-lists; -Wl,, -fplugin, --specs=, @, and =/ are unconditionally blocked.
4. Unbounded request size: internal/api/server.go caps HTTP body at 1 MiB; internal/validator/validator.go bounds source, stdin, and test count.
5. UID collisions: internal/jail/template.go and internal/jail/execute.go rely on namespace isolation (clone_newns, clone_newpid, clone_newipc) rather than unique UIDs.
6. Unbounded output: internal/jail/output.go caps stdout/stderr at 64 KiB per stream per job.
7. Stale directories: internal/runner/run.go defers cleanup on every job; internal/runner/orphans.go runs a background sweeper.

Languages

Python 3 and C++ with full build and test support. Adding a language requires only a YAML entry in configs/languages.yaml with no Go code changes.

Testing

E2E tests in tests/e2e_test.go exercise the full request lifecycle: build, run, status classification, and cleanup. Run with make integration. Pre-submission verification in testdata/check.sh (make check) validates bounded queuing, API responses, status codes, and error envelopes across all scenarios.

[thesouldev]

Evaluation Summary

Team: pr28-HUGO · Rank: 7 / 45 · Weighted score: 67.7 / 100

Area (weight)	Score	What it covers
Technical (~60%)	82.8%	API contract, security holes, concurrency + benchmarks, plug-and-play languages
Code quality / SDLC (~30%)	42.9%	tests, clean git history, lint, docs
Communication (~10%)	50.0%	README clarity, framework justification

Checks:

• Build: Yes
• Health: Yes
• CI: no-actions
• Sandbox: nsjail
• Happy-path: 9/9

Thanks for participating! Scores reflect evaluation against the spec. Reply here if you have questions.