fix(cred_proxy): close git-push bypass + route through pipelock (PRD 0010)
test / unit (pull_request) Successful in 15s
test / integration (pull_request) Successful in 29s

Three coupled fixes that close a documented bypass of git-gate's
gitleaks pre-receive hook:

1. cred-proxy refuses git smart-HTTP push at runtime. Any path
   ending in /git-receive-pack or /info/refs?service=git-receive-pack
   returns 403 with a pointer at the bottle.git SSH path. Fetch
   (upload-pack) is still allowed — the bypass we're closing is
   push, where gitleaks is the load-bearing scanner. Hard guarantee.

2. The provisioner suppresses the cred-proxy `~/.gitconfig` insteadOf
   rewrite for any host already declared in bottle.git. git-gate is
   the canonical git path there; we don't write a competing rule
   that would let `git clone https://<host>/...` succeed in ways
   that confuse on push. Defense in depth — (1) is the hard guarantee.

3. cred-proxy routes its outbound HTTPS through pipelock. The
   sidecar's environ now sets HTTPS_PROXY=<pipelock-url>, and the
   image's entrypoint runs `update-ca-certificates` over the
   per-bottle pipelock CA (docker cp'd into
   /usr/local/share/ca-certificates/pipelock.crt before start) so
   the proxy's HTTPS client trusts pipelock's bumped certs.

   Consequence: pipelock's allowlist + body scanner now sit in the
   cred-proxy egress path the same way they sit in front of direct
   agent traffic. The cred-proxy upstream hosts (api.github.com,
   github.com, gitea hosts, registry.npmjs.org) come OFF
   pipelock's passthrough_domains. Only api.anthropic.com remains
   on passthrough (LLM body content legitimately trips DLP).

PRD 0010 updated to reflect all three. Tests adjusted: the
"cred-proxy hosts go on passthrough" assertion in
test_pipelock_allowlist flips to "they don't", a new
TestIsGitPushRequest exercises the smart-HTTP refusal predicate,
and the gitconfig renderer tests cover the per-host suppression
matrix.
This commit is contained in:
2026-05-13 21:09:33 -04:00
parent c8ab90d01d
commit 27b2d78b11
12 changed files with 329 additions and 63 deletions
+15 -5
View File
@@ -64,16 +64,24 @@ class CredProxyPlan:
The slug + routes_path + upstreams + token_env_map fields are
filled at prepare time (host-side, side-effect-free on docker).
The network fields are populated by the backend's launch step
via `dataclasses.replace` once those networks exist. Empty
defaults are sentinels meaning "not yet set"; `.start` validates
that they are populated.
The network + pipelock fields are populated by the backend's
launch step via `dataclasses.replace` once those resources
exist. Empty defaults are sentinels meaning "not yet set";
`.start` validates that they are populated.
`token_env_map` is `{<token_env in container>: <TokenRef on host>}`.
The backend's start step reads `os.environ[TokenRef]` and forwards
the value into the cred-proxy container's environ under
`token_env`. The plan itself never holds token values — secrets
never land in a dataclass that might be logged."""
never land in a dataclass that might be logged.
`pipelock_ca_host_path` is the host path of the per-bottle CA
pipelock will present on bumped TLS handshakes; the cred-proxy
image's entrypoint runs `update-ca-certificates` over it so the
proxy's HTTPS client trusts pipelock's CA. `pipelock_proxy_url`
is the URL cred-proxy sets as `HTTPS_PROXY` in its environ so
outbound HTTPS traverses pipelock — making pipelock's body
scanner part of the cred-proxy egress path."""
slug: str
routes_path: Path
@@ -81,6 +89,8 @@ class CredProxyPlan:
token_env_map: dict[str, str]
internal_network: str = ""
egress_network: str = ""
pipelock_ca_host_path: Path = Path()
pipelock_proxy_url: str = ""
# Hardcoded upstream URLs for the non-gitea Kinds. Gitea's URL is per-