mirror of
https://github.com/pewdiepie-archdaemon/odysseus.git
synced 2026-09-16 13:12:20 +02:00
fix(skills): harden skill import against DNS rebinding and SSRF TOCTOU (#5986)
* fix(skill-importer): validate URL scheme and improve skills.sh handling * fix(skill-importer): enhance DNS resolution and SSRF protection in fetch URL handling * fix(url-safety): add allowed_dist parameter to check_outbound_url for flexible private blocking * test(skill-importer): add comprehensive tests for URL parsing and outbound checks * ensure newline at end of file in test_check_outbound_url_allows_public_ip * fix(skill-importer): improve TLS certificate handling in _get_checked function * fix(skill-importer): enhance _check_fetch_url to handle both hostnames and full URLs * fix(skill-importer): enhance parse_skill_source to support skills.sh URLs in path and netloc * fix(skill-importer): simplify skills.sh hostname check in parse_skill_source * fix(skill-importer): enhance parse_skill_source to identify skills.sh URLs in path and handle localhost/IP addresses * fix(skill-importer): enhance _resolve_and_check_url to validate all resolved IP addresses and prevent TOCTOU vulnerabilities * fix(skill-importer): enhance parse_skill_source to support schemeless GitHub and skills.sh URLs * fix(memory): resolve CodeQL URL sanitization warning and restore _check_fetch_url test alias * fix(memory): pin skill fetch sockets without rewriting URLs * fix(memory): reject unsupported skill wrapper hosts * refactor(url-safety): remove unused importer exception * test(memory): keep redirect regression hermetic * test(dns-rebinding): add test for _PinnedTransport to ensure connection to pinned IP * fix(skill-importer): enhance skills.sh support to extract GitHub links from page content * fix(skill-importer): improve URL scheme validation for GitHub and skills.sh links * fix(skills): reject unusable skill URLs instead of guessing Resolving a skills.sh link by scraping the first github.com URL out of the page body cannot work. Skill pages only ever link the repository root, never the skill's subdirectory, so every skill in a repo resolved to the same bundle: importing skills.sh/anthropics/skills/pdf walked the whole monorepo, saturated the 64-file cap, and installed algorithmic-art behind an ok:true response. Restore the redirect-target unwrap and fail with a message that says what to do instead. Also report the real reason a URL is rejected. The scheme check keyed off "://" appearing anywhere in the string, so a supplied-but-unusable URL came back as "URL is required", and a schemeless URL carrying "://" in its query was reported as an unsupported scheme. Key off the parsed scheme and let opaque schemes (mailto:, javascript:) and a schemeless host:port fall through to the host check. * test(skills): tighten the real-socket pinning regression The handler swallowed its own exceptions, so a failure inside it surfaced as a confusing assertion on the captured client address. Record the exception and assert on it, run the thread as a daemon, and close the listening socket from the test so a hang cannot outlive the run. Also drop the duplicate ipaddress import and the missing newline. * fix(skills): require exact GitHub skill URLs * test(skills): read complete pinned request headers --------- Co-authored-by: RaresKeY <158580472+RaresKeY@users.noreply.github.com> Co-authored-by: Léo <leograndcontact@gmail.com>
This commit is contained in:
Boody
@@ -1,16 +1,18 @@
|
||||
"""Import SKILL.md bundles from public GitHub (or skills.sh → GitHub) URLs."""
|
||||
from __future__ import annotations
|
||||
|
||||
import ipaddress
|
||||
import logging
|
||||
import os
|
||||
import re
|
||||
import time
|
||||
from dataclasses import dataclass
|
||||
from typing import Dict, List, Optional, Tuple
|
||||
from typing import Dict, Iterable, List, Optional, Tuple, cast
|
||||
from urllib.parse import quote, urljoin, urlparse
|
||||
|
||||
import httpcore
|
||||
import httpx
|
||||
|
||||
from src.url_safety import check_outbound_url
|
||||
from src.url_safety import _default_resolver, check_outbound_url
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
@@ -25,6 +27,7 @@ TEXT_NAMES = {"skill.md", "license", "license.md", "readme.md"}
|
||||
_GITHUB_HOSTS = frozenset({
|
||||
"github.com", "www.github.com", "api.github.com", "raw.githubusercontent.com",
|
||||
})
|
||||
_SKILLS_SH_HOSTS = frozenset({"skills.sh", "www.skills.sh"})
|
||||
|
||||
|
||||
def _github_host(url: str) -> str:
|
||||
@@ -72,18 +75,158 @@ def _is_text_file(name: str) -> bool:
|
||||
_MAX_FETCH_REDIRECTS = 5
|
||||
|
||||
|
||||
def _check_fetch_url(url: str) -> None:
|
||||
"""SSRF guard for skill-import fetches (defense-in-depth).
|
||||
def _validated_ips(raw_ips: List[str]) -> List[ipaddress._BaseAddress]:
|
||||
"""Parse and de-duplicate one resolver snapshot in resolver order."""
|
||||
ips: List[ipaddress._BaseAddress] = []
|
||||
seen = set()
|
||||
for raw in raw_ips:
|
||||
if not isinstance(raw, str):
|
||||
continue
|
||||
try:
|
||||
ip = ipaddress.ip_address(raw.split("%", 1)[0])
|
||||
except ValueError:
|
||||
continue
|
||||
if ip in seen:
|
||||
continue
|
||||
seen.add(ip)
|
||||
ips.append(ip)
|
||||
return ips
|
||||
|
||||
Skill bundles only ever come from public GitHub, never an internal
|
||||
address, so block private/loopback/link-local targets on every hop —
|
||||
matching the hardened web-fetch path in
|
||||
``services/search/content.py:_get_public_url`` rather than the lenient
|
||||
default used for admin-configured model endpoints.
|
||||
"""
|
||||
ok, reason = check_outbound_url(url, block_private=True)
|
||||
|
||||
def _resolve_and_check_url(url: str) -> List[ipaddress._BaseAddress]:
|
||||
"""Return the exact address snapshot approved for one fetch hop."""
|
||||
resolved_ips: List[str] = []
|
||||
|
||||
def _recording_resolver(host: str) -> List[str]:
|
||||
answers = list(_default_resolver(host))
|
||||
resolved_ips[:] = answers
|
||||
return answers
|
||||
|
||||
ok, reason = check_outbound_url(
|
||||
url,
|
||||
block_private=True,
|
||||
resolver=_recording_resolver,
|
||||
)
|
||||
if not ok:
|
||||
raise SkillImportError(reason)
|
||||
raise SkillImportError(f"outbound URL blocked: {reason}")
|
||||
|
||||
pinned_ips = _validated_ips(resolved_ips)
|
||||
if not pinned_ips:
|
||||
raise SkillImportError("outbound URL blocked: host did not resolve to a usable address")
|
||||
return pinned_ips
|
||||
|
||||
|
||||
# Backward compatibility alias for tests importing _check_fetch_url directly
|
||||
_check_fetch_url = _resolve_and_check_url
|
||||
|
||||
|
||||
class _PinnedBackend(httpcore.NetworkBackend):
|
||||
"""Connect only to addresses from one validated DNS snapshot."""
|
||||
|
||||
def __init__(self, ips: List[ipaddress._BaseAddress]):
|
||||
self._ips = [str(ip) for ip in ips]
|
||||
self._real = httpcore.SyncBackend()
|
||||
|
||||
def connect_tcp(
|
||||
self,
|
||||
host: str,
|
||||
port: int,
|
||||
timeout: float | None = None,
|
||||
local_address: str | None = None,
|
||||
socket_options=None,
|
||||
):
|
||||
deadline = None if timeout is None else time.monotonic() + timeout
|
||||
last_exc: Optional[Exception] = None
|
||||
for ip in self._ips:
|
||||
remaining = None if deadline is None else max(0.0, deadline - time.monotonic())
|
||||
try:
|
||||
return self._real.connect_tcp(
|
||||
ip,
|
||||
port,
|
||||
remaining,
|
||||
local_address,
|
||||
socket_options,
|
||||
)
|
||||
except (httpcore.ConnectError, httpcore.ConnectTimeout) as exc:
|
||||
last_exc = exc
|
||||
if deadline is not None and time.monotonic() >= deadline:
|
||||
break
|
||||
if last_exc is not None:
|
||||
raise last_exc
|
||||
raise httpcore.ConnectError("no validated address available")
|
||||
|
||||
def connect_unix_socket(self, path, timeout=None, socket_options=None):
|
||||
return self._real.connect_unix_socket(path, timeout, socket_options)
|
||||
|
||||
def sleep(self, seconds: float) -> None:
|
||||
return self._real.sleep(seconds)
|
||||
|
||||
|
||||
_HTTPCORE_TO_HTTPX_EXC = {
|
||||
httpcore.ConnectError: httpx.ConnectError,
|
||||
httpcore.ConnectTimeout: httpx.ConnectTimeout,
|
||||
httpcore.LocalProtocolError: httpx.LocalProtocolError,
|
||||
httpcore.NetworkError: httpx.NetworkError,
|
||||
httpcore.PoolTimeout: httpx.PoolTimeout,
|
||||
httpcore.ProtocolError: httpx.ProtocolError,
|
||||
httpcore.ProxyError: httpx.ProxyError,
|
||||
httpcore.ReadError: httpx.ReadError,
|
||||
httpcore.ReadTimeout: httpx.ReadTimeout,
|
||||
httpcore.RemoteProtocolError: httpx.RemoteProtocolError,
|
||||
httpcore.TimeoutException: httpx.TimeoutException,
|
||||
httpcore.UnsupportedProtocol: httpx.UnsupportedProtocol,
|
||||
httpcore.WriteError: httpx.WriteError,
|
||||
httpcore.WriteTimeout: httpx.WriteTimeout,
|
||||
}
|
||||
|
||||
|
||||
class _PinnedTransport(httpx.BaseTransport):
|
||||
"""Pin socket connects while preserving URL authority, Host, and TLS SNI."""
|
||||
|
||||
def __init__(self, ips: List[ipaddress._BaseAddress]):
|
||||
self._pinned_ips = list(ips)
|
||||
self._pool = httpcore.ConnectionPool(
|
||||
ssl_context=httpx.create_ssl_context(),
|
||||
http1=True,
|
||||
http2=False,
|
||||
network_backend=_PinnedBackend(ips),
|
||||
)
|
||||
|
||||
def handle_request(self, request: httpx.Request) -> httpx.Response:
|
||||
core_request = httpcore.Request(
|
||||
method=request.method,
|
||||
url=httpcore.URL(
|
||||
scheme=request.url.raw_scheme,
|
||||
host=request.url.raw_host,
|
||||
port=request.url.port,
|
||||
target=request.url.raw_path,
|
||||
),
|
||||
headers=request.headers.raw,
|
||||
content=request.stream,
|
||||
extensions=request.extensions,
|
||||
)
|
||||
core_response = None
|
||||
try:
|
||||
core_response = self._pool.handle_request(core_request)
|
||||
content = b"".join(cast(Iterable[bytes], core_response.stream))
|
||||
except Exception as exc:
|
||||
mapped = _HTTPCORE_TO_HTTPX_EXC.get(type(exc))
|
||||
if mapped is not None:
|
||||
raise mapped(str(exc)) from exc
|
||||
raise
|
||||
finally:
|
||||
if core_response is not None:
|
||||
core_response.close()
|
||||
|
||||
return httpx.Response(
|
||||
status_code=core_response.status,
|
||||
headers=core_response.headers,
|
||||
content=content,
|
||||
extensions=core_response.extensions,
|
||||
)
|
||||
|
||||
def close(self) -> None:
|
||||
self._pool.close()
|
||||
|
||||
|
||||
def _get_checked(
|
||||
@@ -100,49 +243,76 @@ def _get_checked(
|
||||
hand lets us re-validate every hop, closing that blind-SSRF gap.
|
||||
"""
|
||||
current = url
|
||||
with httpx.Client(follow_redirects=False, timeout=timeout) as client:
|
||||
for _ in range(_MAX_FETCH_REDIRECTS + 1):
|
||||
_check_fetch_url(current)
|
||||
for _ in range(_MAX_FETCH_REDIRECTS + 1):
|
||||
pinned_ips = _resolve_and_check_url(current)
|
||||
with httpx.Client(
|
||||
transport=_PinnedTransport(pinned_ips),
|
||||
follow_redirects=False,
|
||||
timeout=timeout,
|
||||
) as client:
|
||||
r = client.get(current, headers=headers)
|
||||
if r.status_code in (301, 302, 303, 307, 308):
|
||||
location = r.headers.get("location")
|
||||
if not location:
|
||||
return r
|
||||
current = urljoin(str(r.url), location)
|
||||
continue
|
||||
return r
|
||||
|
||||
if r.status_code in (301, 302, 303, 307, 308):
|
||||
location = r.headers.get("location")
|
||||
if not location:
|
||||
return r
|
||||
current = urljoin(str(r.url), location)
|
||||
continue
|
||||
return r
|
||||
raise SkillImportError("too many redirects while fetching skill bundle")
|
||||
|
||||
|
||||
def parse_skill_source(url: str) -> ResolvedSource:
|
||||
"""Normalize skills.sh / GitHub web URLs into owner/repo/ref/path."""
|
||||
raw = (url or "").strip()
|
||||
if not raw:
|
||||
url = (url or "").strip()
|
||||
if not url:
|
||||
raise SkillImportError("URL is required")
|
||||
|
||||
# skills.sh often links to GitHub; try to unwrap ?url= or redirect target later.
|
||||
if "skills.sh" in raw and "github.com" not in raw:
|
||||
r = _get_checked(raw, timeout=20.0)
|
||||
# ``urlparse`` only reports an unambiguous scheme when the URL carries the
|
||||
# ``scheme://`` form. Opaque schemes (``mailto:``, ``javascript:``) and a
|
||||
# schemeless ``host:port`` both parse a "scheme" that is not one, so they
|
||||
# fall through to the host check below and are rejected on the host instead.
|
||||
scheme = urlparse(url).scheme.lower()
|
||||
if scheme not in ("http", "https"):
|
||||
if scheme and url.lower().startswith(f"{scheme}://"):
|
||||
raise SkillImportError(f"unsupported URL scheme: {scheme}")
|
||||
# Schemeless "github.com/owner/repo" — accept only a supported host.
|
||||
rough_host = (urlparse("//" + url).hostname or "").lower()
|
||||
if rough_host not in _GITHUB_HOSTS and rough_host not in _SKILLS_SH_HOSTS:
|
||||
raise SkillImportError("Only GitHub or skills.sh URLs are supported")
|
||||
url = "https://" + url
|
||||
|
||||
parsed = urlparse(url)
|
||||
hostname = (parsed.hostname or "").lower()
|
||||
if hostname not in _GITHUB_HOSTS and hostname not in _SKILLS_SH_HOSTS:
|
||||
raise SkillImportError("Only GitHub or skills.sh URLs are supported")
|
||||
|
||||
# A skills.sh link is only usable if it redirects to an exact supported
|
||||
# GitHub host. Scraping the page body for a github.com link cannot work:
|
||||
# skill pages only ever link the repository root, never the skill's
|
||||
# subdirectory, so the scrape resolves every skill in a repo to the same
|
||||
# (wrong) bundle. Fail with an actionable message instead.
|
||||
if hostname in _SKILLS_SH_HOSTS:
|
||||
r = _get_checked(url, timeout=20.0)
|
||||
if r.status_code >= 400:
|
||||
raise _github_response_error(r)
|
||||
final = str(r.url)
|
||||
_assert_github_url(final, context="redirect target")
|
||||
# Page may embed a github link; prefer final URL if redirected.
|
||||
if "github.com" in final:
|
||||
raw = final
|
||||
else:
|
||||
m = re.search(r"https?://github\.com/[^\s\"')]+", r.text or "")
|
||||
if m:
|
||||
raw = m.group(0).rstrip(".,)")
|
||||
if _github_host(final) not in _GITHUB_HOSTS:
|
||||
raise SkillImportError(
|
||||
"skills.sh did not redirect to GitHub — open the skill's "
|
||||
"repository on GitHub, navigate to the exact skill folder or "
|
||||
"SKILL.md file, and paste that URL; the repository-root link "
|
||||
"alone is not sufficient"
|
||||
)
|
||||
url = final
|
||||
|
||||
parsed = urlparse(raw)
|
||||
host = _github_host(raw)
|
||||
if host not in _GITHUB_HOSTS:
|
||||
raise SkillImportError(
|
||||
"Only GitHub URLs are supported (https://github.com/... or raw.githubusercontent.com/...)"
|
||||
)
|
||||
# Update parsed and hostname to reflect the new GitHub URL
|
||||
parsed = urlparse(url)
|
||||
hostname = (parsed.hostname or "").lower()
|
||||
|
||||
if host == "raw.githubusercontent.com":
|
||||
_assert_github_url(url)
|
||||
|
||||
if hostname == "raw.githubusercontent.com":
|
||||
# /owner/repo/ref/path/to/file
|
||||
bits = [p for p in parsed.path.split("/") if p]
|
||||
if len(bits) < 4:
|
||||
|
||||
Reference in New Issue
Block a user