diff --git a/.github/workflows/site-numbers.yml b/.github/workflows/site-numbers.yml
new file mode 100644
index 0000000..138e66c
--- /dev/null
+++ b/.github/workflows/site-numbers.yml
@@ -0,0 +1,68 @@
+# Site numbers come from the claims register in bawbel/.github (scripts/sync_site_numbers.py).
+# On a pull request: fail if the stat block is stale or a count is typed outside it.
+# Weekly and on demand: regenerate the block and open a pull request against develop.
+name: site numbers
+
+on:
+ pull_request:
+ branches: [develop, main]
+ schedule:
+ - cron: "17 6 * * 1"
+ workflow_dispatch:
+
+permissions:
+ contents: read
+
+jobs:
+ check:
+ if: github.event_name == 'pull_request'
+ runs-on: ubuntu-latest
+ timeout-minutes: 10
+ steps:
+ - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4.4.0
+ with:
+ persist-credentials: false
+ - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+ with:
+ python-version: "3.12"
+ - run: python -m pip install -r scripts/requirements.txt
+ - run: python -m pytest tests -q
+ - run: python scripts/sync_site_numbers.py --check
+
+ update:
+ if: github.event_name != 'pull_request'
+ runs-on: ubuntu-latest
+ timeout-minutes: 10
+ permissions:
+ contents: write
+ pull-requests: write
+ steps:
+ - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262 # v4.4.0
+ with:
+ ref: develop
+ - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+ with:
+ python-version: "3.12"
+ - run: python -m pip install -r scripts/requirements.txt
+ - run: python scripts/sync_site_numbers.py
+ - name: Open a pull request when the numbers changed
+ env:
+ GH_TOKEN: ${{ github.token }}
+ run: |
+ if git diff --quiet; then
+ echo "Stat block already current."
+ exit 0
+ fi
+ if [ -n "$(gh pr list --base develop --state open --search 'head:chore/site-numbers-' --json number --jq '.[].number')" ]; then
+ echo "A site-numbers pull request is already open."
+ exit 0
+ fi
+ branch="chore/site-numbers-$(date -u +%Y%m%d)"
+ git config user.name "github-actions[bot]"
+ git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+ git switch -c "$branch"
+ git commit -s -am "chore: sync site numbers from the claims register"
+ git push origin "$branch"
+ gh pr create --base develop --head "$branch" \
+ --title "chore: sync site numbers from the claims register" \
+ --body "Generated by scripts/sync_site_numbers.py from claims/register.yaml in bawbel/.github. Review the new values against that register before merging."
diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..6c56ff1
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,2 @@
+__pycache__/
+.pytest_cache/
diff --git a/README.md b/README.md
index 487fb81..b40cec6 100644
--- a/README.md
+++ b/README.md
@@ -11,7 +11,7 @@ before they reach production. Free forever. Apache 2.0.
[](https://pypi.org/project/bawbel-scanner/)
[](https://pepy.tech/project/bawbel-scanner)
-[](https://github.com/bawbel/ave)
+[](https://github.com/bawbel/ave)
[](https://github.com/bawbel/scanner/blob/main/LICENSE)
@@ -23,7 +23,7 @@ before they reach production. Free forever. Apache 2.0.
| Repo | What it is |
|---|---|
| [bawbel/scanner](https://github.com/bawbel/scanner) | CLI scanner for agentic AI components -- SKILL.md files, MCP servers, system prompts. Confidence and evidence metadata on every finding. |
-| [bawbel/ave](https://github.com/bawbel/ave) | AVE -- the behavioral classification standard for agentic AI components. 51 published records, scored with OWASP AIVSS v0.8, mapped to OWASP MCP Top 10 and MITRE ATLAS. |
+| [bawbel/ave](https://github.com/bawbel/ave) | AVE -- the behavioral classification standard for agentic AI components. Published records scored with OWASP AIVSS v0.8, mapped to OWASP MCP Top 10 and MITRE ATLAS. |
| [bawbel/piranha-api](https://github.com/bawbel/piranha-api) | Free, no-auth REST API serving the full AVE record set and threat intelligence. |
| [bawbel/integrations](https://github.com/bawbel/integrations) | GitHub Action, VS Code extension, and pre-commit hooks. |
@@ -46,6 +46,17 @@ Your CI pipeline scans dependencies for known package vulnerabilities. It does n
---
+### Site numbers
+
+Every count on bawbel.io comes from the claims register in
+[bawbel/.github](https://github.com/bawbel/.github/blob/main/claims/register.yaml), which that
+repository's verifier checks against the code. `scripts/sync_site_numbers.py` writes the values
+into the stat block in `index.html`. A pull request fails when the block is stale or a count is
+typed anywhere else, and a weekly workflow opens a pull request when the register changes. To
+change a number, change the register.
+
+---
+
### Links
[bawbel.io](https://bawbel.io) ยท
diff --git a/docs/index.html b/docs/index.html
index baadd61..bdcaeba 100644
--- a/docs/index.html
+++ b/docs/index.html
@@ -434,12 +434,12 @@
What is covered
-
121 detection rules
-
Pattern, YARA, and Semgrep rules mapped to all 51 AVE records. Every finding now carries confidence and evidence metadata.
+
Detection rules
+
Pattern, YARA, and Semgrep rules, each tied to the AVE record it detects. Every finding now carries confidence and evidence metadata.
Individual findings that are each only MEDIUM severity can combine into a CRITICAL attack chain. 12 built-in chains detect exactly this -- for example, a credential-read finding plus a data-exfiltration finding in the same component becomes a single Credential Exfiltration Chain at CRITICAL 9.8, rather than two separate MEDIUM findings a reviewer might triage as low priority individually.
+
Individual findings that are each only MEDIUM severity can combine into a CRITICAL attack chain. Built-in chains detect exactly this -- for example, a credential-read finding plus a data-exfiltration finding in the same component becomes a single Credential Exfiltration Chain at CRITICAL 9.8, rather than two separate MEDIUM findings a reviewer might triage as low priority individually.
Chain confidence is the minimum confidence across all contributing findings -- the weakest link in the chain, not the average. This is deliberate: a chain is only as trustworthy as its least certain component.
-
Use bawbel chain to filter specifically to delegation-chain findings (AVE-2026-00048), or run a full bawbel scan to see all 12 toxic flow chains evaluated together.
+
Use bawbel chain to filter specifically to delegation-chain findings (AVE-2026-00048), or run a full bawbel scan to see every toxic flow chain evaluated together.
12 built-in attack chains. Credential-read plus data-exfil becomes CRITICAL 9.8 automatically. Combined AIVSS score and minimum confidence across the full chain.
+
Built-in attack chains. Credential-read plus data-exfil becomes CRITICAL 9.8 automatically. Combined AIVSS score and minimum confidence across the full chain.
-
6 Detection Engines
-
Pattern (40 rules), YARA (39), Semgrep (41), LLM semantic, Magika content-type, Docker sandbox. 121 rules total across 51 AVE records.
+
Detection Engines
+
Pattern, YARA, and Semgrep rules, LLM semantic analysis, Magika content-type checks, and a Docker sandbox stage that is off by default. Each rule is tied to the AVE record it detects.
@@ -450,7 +451,7 @@
Maps to the frameworks you already report agai
Required on every record
OWASP MCP Top 10
-
owasp_mcp is required on every one of the 51 AVE records, MCP01 through MCP10. Every finding your scan produces lands in a category your security team already tracks.
+
owasp_mcp is required on every AVE record, MCP01 through MCP10. Every finding your scan produces lands in a category your security team already tracks.
Mapped where applicable
@@ -484,24 +485,28 @@
How Bawbel compares
-
Feature
-
Bawbel
-
Snyk agent-scan
-
ClawGuard
-
Cisco DefenseClaw
+
Feature
+
Bawbel
+
NVIDIA SkillSpector
+
Snyk agent-scan
+
Cisco DefenseClaw
-
Executes MCP servers during scan
Never
Yes
No
Sandboxed
-
Open behavioral classification standard
Yes (51 records)
No
No
No
-
OWASP AIVSS v0.8 scores
Yes
No
No
No
-
Confidence and evidence metadata
Yes
No
No
No
-
Toxic flow detection
Yes (12 chains)
No
No
No
-
Justified suppression with expiry
Yes
No
No
No
-
Conformance grading
Yes (A+ to F)
No
No
No
-
License
Apache 2.0
Apache 2.0
MIT
Proprietary
+
Starts MCP servers during a scan
Never
Never
Yes (stdio servers)
?
+
Runs scanned skill content
Only in the optional Docker sandbox stage, off by default
? means not verifiable from each project's public README at the time of writing (2026-09-12). The same table, with its sources, is in the scanner README.
Free, no auth, Apache 2.0. Every scanner finding includes a piranha_url linking to the full record with IOCs and remediation.
-
GET
/records
all 51 AVE records, filterable
+
GET
/records
all AVE records, filterable
GET
/records/{ave_id}
full record with IOCs and AIVSS
GET
/registry-scan/latest
latest Smithery scan results
GET
/stats/ecosystem
aggregate stats across all sources
diff --git a/scripts/requirements.txt b/scripts/requirements.txt
new file mode 100644
index 0000000..f3d4305
--- /dev/null
+++ b/scripts/requirements.txt
@@ -0,0 +1,2 @@
+pyyaml>=6,<7
+pytest>=8,<10
diff --git a/scripts/sync_site_numbers.py b/scripts/sync_site_numbers.py
new file mode 100644
index 0000000..4b8a141
--- /dev/null
+++ b/scripts/sync_site_numbers.py
@@ -0,0 +1,181 @@
+#!/usr/bin/env python3
+"""Render the site's numbers from the claims register.
+
+Every count on bawbel.io comes from one place: the `expect` values in
+claims/register.yaml in bawbel/.github. That repository's verifier checks each value
+against the code. This script writes them into the stat block of index.html, between
+the stat:start and stat:end markers, and a count typed anywhere else fails --check.
+
+ python scripts/sync_site_numbers.py # rewrite the stat block
+ python scripts/sync_site_numbers.py --check # exit 1 if the block is stale or a
+ # count is typed outside it
+
+The register is read from --register (a file path or an https URL), else from the
+SITE_CLAIMS_REGISTER environment variable, else from REGISTER_URL_DEFAULT.
+
+Exit status: 0 current (or rewritten), 1 stale or a typed count (--check only), 2 the
+register or the stat block could not be read. Nothing is written when the exit is 2.
+"""
+
+from __future__ import annotations
+
+import argparse
+import os
+import re
+import sys
+import urllib.request
+from pathlib import Path
+from urllib.parse import urlsplit
+
+import yaml
+
+ROOT = Path(__file__).resolve().parent.parent
+INDEX = ROOT / "index.html"
+PAGES = (ROOT / "index.html", ROOT / "docs" / "index.html", ROOT / "README.md")
+
+REGISTER_URL_DEFAULT = "https://raw.githubusercontent.com/bawbel/.github/main/claims/register.yaml"
+REGISTER_ENV = "SITE_CLAIMS_REGISTER"
+REGISTER_SCHEMA = "bawbel/claims-register/v1"
+FETCH_TIMEOUT_SEC = 15
+MAX_REGISTER_BYTES = 1024 * 1024
+
+# The stat block in display order: (claim id in the register, label). The register's
+# appears_in patterns match these labels, so a label change needs a register change.
+STATS = (
+ ("C-01", "AVE Records published"),
+ ("C-03", "With a detection engine"),
+ ("C-06", "Attack Chains"),
+ ("C-08", "Detection Engines"),
+ ("C-07", "Detection Rules"),
+)
+
+START = ""
+END = ""
+BLOCK = re.compile(r"(?P[ \t]*)\n.*?", re.S)
+
+# A count typed into prose: "51 AVE records", "121 rules total", "6 Detection Engines".
+# One space only: sample output such as "Findings: 2 Chains: 1" uses two.
+TYPED_COUNT = re.compile(
+ r"(?i)\b\d+ (?:(?:published|built-in|AVE|detection|toxic flow|attack|capability) )*"
+ r"(?:records?|rules?|engines?|chains?|endpoints?)\b"
+)
+# A count in a shields.io badge: "AVE_records-51".
+BADGE_COUNT = re.compile(r"(?i)(? blocks are not checked.
+PRE = re.compile(r"", re.S | re.I)
+
+
+class RegisterError(Exception):
+ """The register could not be read, or lacks a value the site needs."""
+
+
+def read_register(source: str) -> dict:
+ """The parsed register from a file path or an https URL."""
+ if os.path.isfile(source):
+ text = Path(source).read_text(encoding="utf-8")
+ else:
+ parts = urlsplit(source)
+ if parts.scheme != "https" or not parts.netloc:
+ raise RegisterError(f"not a file or an https URL: {source}")
+ req = urllib.request.Request(source, headers={"User-Agent": "bawbel-site-numbers"})
+ try:
+ # B310: the scheme is checked above, so only https reaches urlopen.
+ with urllib.request.urlopen(req, timeout=FETCH_TIMEOUT_SEC) as resp: # nosec B310
+ body = resp.read(MAX_REGISTER_BYTES + 1)
+ except OSError as exc: # URLError, HTTPError, and timeouts are all OSError
+ raise RegisterError(f"{source}: {exc}") from exc
+ if len(body) > MAX_REGISTER_BYTES:
+ raise RegisterError(f"{source}: larger than {MAX_REGISTER_BYTES} bytes")
+ text = body.decode("utf-8", errors="replace")
+ try:
+ data = yaml.safe_load(text)
+ except yaml.YAMLError as exc:
+ raise RegisterError(f"{source}: not YAML: {exc}") from exc
+ if not isinstance(data, dict) or data.get("schema") != REGISTER_SCHEMA:
+ raise RegisterError(f"{source}: not a {REGISTER_SCHEMA} document")
+ return data
+
+
+def stat_values(register: dict) -> dict[str, int]:
+ """The expect value of every claim in STATS. Each must be a whole number."""
+ claims = {c.get("id"): c for c in register.get("claims") or [] if isinstance(c, dict)}
+ values = {}
+ for claim_id, _ in STATS:
+ expect = (claims.get(claim_id) or {}).get("expect")
+ if isinstance(expect, bool) or not isinstance(expect, int) or expect < 0:
+ raise RegisterError(f"{claim_id}: expect is not a whole number: {expect!r}")
+ values[claim_id] = expect
+ return values
+
+
+def render_block(values: dict[str, int], indent: str) -> str:
+ lines = [indent + START]
+ for claim_id, label in STATS:
+ lines.append(
+ f'{indent}
{values[claim_id]}
'
+ f'
{label}
'
+ )
+ lines.append(indent + END)
+ return "\n".join(lines)
+
+
+def typed_counts(path: Path) -> list[str]:
+ """Counts typed into a page outside the stat block and
samples."""
+
+ def blank(match: re.Match) -> str:
+ return "\n" * match.group(0).count("\n") # keep line numbers
+
+ text = PRE.sub(blank, BLOCK.sub(blank, path.read_text(encoding="utf-8")))
+ return [
+ f"{path.relative_to(ROOT)}:{lineno}: {match.group(0)!r}"
+ for lineno, line in enumerate(text.splitlines(), 1)
+ for pattern in (TYPED_COUNT, BADGE_COUNT)
+ for match in pattern.finditer(line)
+ ]
+
+
+def main(argv: list[str] | None = None) -> int:
+ parser = argparse.ArgumentParser(description=__doc__.splitlines()[0])
+ parser.add_argument("--check", action="store_true",
+ help="change nothing; exit 1 if the block is stale or a count is typed")
+ parser.add_argument("--register", default=os.environ.get(REGISTER_ENV) or REGISTER_URL_DEFAULT,
+ help="claims register: a file path or an https URL")
+ args = parser.parse_args(argv)
+
+ try:
+ values = stat_values(read_register(args.register))
+ except RegisterError as exc:
+ print(f"sync_site_numbers: {exc}", file=sys.stderr)
+ return 2
+
+ html = INDEX.read_text(encoding="utf-8")
+ match = BLOCK.search(html)
+ if match is None:
+ print("sync_site_numbers: index.html has no stat:start/stat:end block", file=sys.stderr)
+ return 2
+ updated = html[: match.start()] + render_block(values, match.group("indent")) + html[match.end():]
+ summary = ", ".join(f"{claim_id} {values[claim_id]}" for claim_id, _ in STATS)
+
+ if args.check:
+ problems = []
+ if updated != html:
+ problems.append("index.html: the stat block is stale; run scripts/sync_site_numbers.py")
+ for page in PAGES:
+ problems += [f"{hit}: a count typed outside the stat block" for hit in typed_counts(page)]
+ for problem in problems:
+ print(problem, file=sys.stderr)
+ if problems:
+ return 1
+ print(f"site numbers match the claims register: {summary}")
+ return 0
+
+ if updated == html:
+ print(f"index.html: stat block already current: {summary}")
+ else:
+ INDEX.write_text(updated, encoding="utf-8")
+ print(f"index.html: stat block updated: {summary}")
+ return 0
+
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests/test_sync_site_numbers.py b/tests/test_sync_site_numbers.py
new file mode 100644
index 0000000..27f84b2
--- /dev/null
+++ b/tests/test_sync_site_numbers.py
@@ -0,0 +1,111 @@
+"""Tests for scripts/sync_site_numbers.py. Run: python -m pytest tests -q"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+import pytest
+import yaml
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent / "scripts"))
+import sync_site_numbers as sync # noqa: E402
+
+VALUES = {"C-01": 87, "C-03": 43, "C-06": 12, "C-07": 120, "C-08": 6}
+
+PAGE = """
+