SiteProbe v0.3.0

SiteProbe is now native Kujo throughout: crawl policy, robots, extraction, analysis, validation, signing, comparison and reporting. Python is only a development/test dependency.

• Preserves the existing v1 artifact schemas, CLI commands and safe defaults.
• Adds bounded streaming for large artifacts and sitemaps, stricter trust boundaries and immutable publication.
• Performs real bounded concurrent fetching with isolated workers.
• Uses a pinned Kujo runtime with loop/exception corrections, bounded regex reuse and Windows publication fixes.
• Qualifies Linux, macOS and Windows, with complete 10,000-page native crawls at concurrency 1/4/8/16.

The archives are platform-qualified source packages, not standalone binaries. They require Kujo revision 2be1f04b89dbecd591ef1af0d68974c05586900e; set KUJO_BIN to that runtime. Python is unnecessary for product commands. Every ZIP has a SHA-256 checksum file. On Unix, use bash siteprobe if extraction does not preserve executable permissions.

See the packaged README and release qualification for installation and verification. The 0.2.0 changelog section records an unpublished development milestone; this release includes all changes since v0.1.0.

SiteProbe 0.1.0

SiteProbe 0.1.0 release qualification

Status: PASS for the documented bounded, read-only crawler contract.

The release gate covers deterministic URL fuzzing; malformed HTML and JSON-LD;
redirect chains and cross-origin redirect rejection; robots and canonical
conflicts; duplicate content; cyclic links; timeouts, bounded retries, 429/5xx
handling; response/page/depth/concurrency/output/report budgets; stale output
rejection; offline command boundaries; and semantic deterministic reruns.
Fixture servers record request methods and the suite fails if SiteProbe uses
anything other than GET.

The WebOps dogfood baseline is
kujo-workflows/docs/evidence/siteprobe-live-final-2026-08-11: 60 public pages,
1,177 links, and zero findings. Release dogfood must use the same 60-page scope
and compare the resulting run to that immutable baseline.

Run bash scripts/validate.sh and python3 scripts/benchmark.py --pages 1000.
SiteProbe has no publish, submit, or ACT command.