feat(supervisor-middleware): add network egress middleware by pimlock · Pull Request #2027 · NVIDIA/OpenShell

pimlock · 2026-06-26T20:30:32Z

Summary

Implements the first usable RFC 0009 supervisor middleware slice: proto-backed, host-selected HTTP egress middleware for HttpRequest/pre_credentials, with both in-process built-ins and statically registered operator-run gRPC services.

The implementation covers RFC 0009 Phase 1 and adds basic external-service support from Phase 2. It establishes the contract, policy plumbing, ordered chain execution, built-in secret redaction, static gateway registration, relay integration, validation before policy persistence, body limits, audit events, and user-facing configuration and operations documentation.

Tip

See example middleware implementation: #2169

Related Issue

Closes #2010
Part of #1733
Design/RFC: #1738

Changes

Add the supervisor middleware contract and runtime, including the built-in openshell/secrets redactor and statically registered operator-run gRPC services.
Add ordered, host-selector-based network_middlewares policy configuration and validation, independent of the network policy rule that admits a request.
Enforce middleware on inspected HTTP requests before credential injection, with transformations, safe header additions, body limits, and explicit fail-open/fail-closed behavior.
Deliver effective service configuration to supervisors with resilient registry reload and last-known-good behavior.
Add OCSF observability plus gateway, policy, operations, architecture, and extensibility documentation.

Testing

mise run pre-commit passes
Unit and integration-style tests added or updated
E2E coverage considered
- No separate gateway E2E was added. The request relay, policy validation, registry, and remote gRPC paths are covered by automated unit and integration-style tests.

Checklist

Follows Conventional Commits
Commits are signed off (DCO)
Architecture docs updated
Published docs updated

copy-pr-bot · 2026-06-26T20:30:35Z

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

github-actions · 2026-06-30T19:48:46Z

🌿 Preview your docs: https://nvidia-preview-pr-2027.docs.buildwithfern.com/openshell

pimlock · 2026-07-02T21:20:43Z

/ok to test c4b0dcf

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

…are outages An unreachable operator-registered middleware service previously aborted sandbox startup via a hard error in load_policy, contradicting the per-request on_error contract and the resilient live-reload path. Retry the initial connect and, on failure, degrade to the built-in registry so matched requests are governed by each config's on_error (deny for fail_closed, allow for fail_open) instead of blocking the whole sandbox. The policy poll loop now reconciles the registry on every poll while an install is pending, so a recovered service is adopted without waiting for a config change; a failed reconcile also no longer blocks unrelated policy updates. Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

…limit A chain entry whose binding did not resolve reported a zero body limit, which dragged the whole chain's buffer cap to zero and spuriously failed body-bearing requests over capacity even when a resolved middleware could have processed them. Exclude unresolved entries from the limit via a new DescribedChainEntry::is_resolved(); when no entry resolves, skip buffering and apply each entry's on_error directly. Also fix two parallel-test flakes found while validating the change: - Build middleware OCSF events into a Vec and assert on it directly instead of capturing through the global tracing pipeline, whose callsite-interest cache is process-global and raced under parallel runs. - Accumulate the websocket deny response until the reason marker arrives rather than assuming a single read returns the full body. Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

github-actions · 2026-07-03T20:54:13Z

Label test:e2e applied for 3b58759. Open the existing run and click Re-run all jobs to execute with the label set. The run will execute the standard E2E suite after building the required gateway and supervisor images once. The matching required CI gate status on this PR will flip green automatically once the run finishes.

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

TaylorMutch · 2026-07-07T21:49:06Z

+Plan startup and updates around these boundaries:
+
+- Start registered services before the gateway. The gateway validates every registration during startup.
+- Keep service endpoints reachable from both the gateway and sandbox supervisors. The supervisors call operator-run services directly on the request path.


Clarifying question - What is the behavior when a service endpoint is unexpectedly or intermittently unavailable? This would be in the case where a on gateway and supervisor startups, the service was healthy, but then becomes unavailable later. What would happen?

This depends on the on_error setting. If it's fail_open, the request will be sent upstream, if it's fail_closed (default), it will be rejected.

TaylorMutch · 2026-07-07T21:51:01Z

+// HttpRequestTarget describes the admitted HTTP destination and request target.
+message HttpRequestTarget {


Again just out of curiosity - are generic headers in scope here or part of the target?

The headers are currently modelled as part of the request itself, this HttpRequestTarget is supposed to represent "where" and the rest (the "what") is in the HttpRequestEvaluation.

The reason for that split was for HttpRequestTarget to represent what the network policy operates on when making its allow/deny decision.

elezar · 2026-07-08T13:51:36Z

+///
+/// Matching is case-insensitive. Invalid or empty patterns return an error
+/// instead of silently becoming a non-match.
+pub fn host_matches(pattern: &str, host: &str) -> std::result::Result<bool, String> {


Nit: Although we're using this to match a host, this is matching a general pattern to a string should we rename the function?

I'm also wondering if it make sense to introduce a Pattern type that wraps glob::Pattern and performs pattern validation on new?

elezar · 2026-07-08T13:54:59Z

+use miette::{Result, miette};
+
+/// Binding identifier for the built-in secret redaction middleware.
+pub const BUILTIN_SECRETS: &str = "openshell/secrets";


If this were go, I would consider implementing this as a type that implements some "Middleware" interface. Is something similar possible / idomatic in Rust?

elezar · 2026-07-08T13:57:04Z

+fn validate_secrets_config(config: &prost_types::Struct) -> Result<()> {
+    let mode = config
+        .fields
+        .get("secrets")
+        .and_then(|value| match value.kind.as_ref() {
+            Some(prost_types::value::Kind::StringValue(value)) => Some(value.as_str()),
+            _ => None,
+        })
+        .unwrap_or("redact");
+    if mode != "redact" {
+        return Err(miette!(
+            "{BUILTIN_SECRETS} only supports config.secrets: redact in phase 1"
+        ));
+    }
+    Ok(())
+}


It is not clear from this function which fields the openshell/secrets config actually has. Does this further justify a concrete type? (Something like what we do for the driver-config structs?)

elezar · 2026-07-08T13:58:20Z

+            Some(prost_types::value::Kind::StringValue(value)) => Some(value.as_str()),
+            _ => None,
+        })
+        .unwrap_or("redact");


So the default value is redact even if no secrets field is present?

elezar · 2026-07-08T13:59:04Z

+
+    #[test]
+    fn secrets_config_defaults_to_redact() {
+        validate_builtin_config(BUILTIN_SECRETS, &prost_types::Struct::default()).unwrap();


I don't actually see redact mentioned in this test at all.

drew

Feedback from agent review

drew · 2026-07-08T14:28:42Z

+                if endpoint.tls == "skip"
+                    && selector_matches_host(middleware, &endpoint.host).unwrap_or(false)
+                {


[P1] Detect tls-skip conflicts against wildcard endpoints

When a tls: skip endpoint uses a wildcard host such as *.example.com and a middleware selector includes a concrete host such as api.example.com, this compares the selector against the literal endpoint pattern, so validation passes even though real traffic for api.example.com is admitted without TLS inspection and the required middleware cannot run. The conflict check needs to reason about selector/endpoint pattern overlap, not only whether the selector matches the endpoint string.

johntmyers · 2026-07-08T16:04:25Z

+            let chain = engine.query_middleware_chain(&middleware_network_input(ctx))?;
+            let req = match apply_middleware_chain(
+                req,
+                client,
+                ctx,
+                chain,
+                engine.middleware_runner(),
+                engine.generation_guard(),
+            )


It looks like L7 policy is evaluated before middleware. So I'd imagine that the middleware could modify or replace an allowed operation with one that would otherwise be denied by policy. Is this intended? If so we should clearly document this.

Great point.

As discussed - we want the policy to be guaranteed by openshell, so we need to add a check after any middleware hook (at least if the body/headers were modified by the middleware). I will make this change.

johntmyers

Agent review of the current head. The blocking enforcement and credential-handling concerns are posted inline. I recommend addressing the P1 findings before merge. Code-only review; tests were not run.

johntmyers · 2026-07-08T19:26:49Z

                        &mut tls_upstream,
                        &ctx,
                        &generation_guard,
+                        Some(&opa_engine),


[P1] Matching middleware still does not make CONNECT inspection mandatory. should_inspect_l7 above is derived only from the endpoint L7 route. For an L4-only endpoint with a matching fail-closed chain, an unsupported HTTP method or h2c preface can take the raw-copy path without invoking middleware. Please include the selected middleware chain when deciding whether inspection is required and fail closed for uninspectable protocols.

johntmyers · 2026-07-08T19:26:49Z


        if allowed || (config.enforcement == EnforcementMode::Audit && !force_deny) {
+            let chain = engine.query_middleware_chain(&middleware_network_input(ctx))?;
+            let req = match apply_middleware_chain(


[P1] Body-aware policy has already evaluated the original JSON-RPC request before this middleware call. An ALLOW result can replace that body with a locally denied operation, and the replacement is forwarded below without reparsing or reevaluation. Please run transformations before body-aware policy evaluation, or reparse and reevaluate the final body.

johntmyers · 2026-07-08T19:26:49Z

+            };
+            for endpoint in &rule.endpoints {
+                if endpoint.tls == "skip"
+                    && selector_matches_host(middleware, &endpoint.host).unwrap_or(false)


[P1] endpoint.host may itself be a wildcard pattern, so treating it as the concrete host argument does not detect selector-pattern intersection. For example, tls: skip on *.example.com and middleware selecting api.example.com pass validation even though runtime traffic is uninspectable. Please conservatively detect pattern overlap; the corresponding runtime validation needs the same fix.

johntmyers · 2026-07-08T19:26:49Z

+        // operator service is picked up without waiting for a policy change.
+        // Failure preserves the last-known-good registry and does not block
+        // the remaining config updates below.
+        reconcile_middleware_registry(


[P1] The registry is reconciled before the new OPA policy is committed. Concurrent requests can therefore evaluate the old policy against the new registry. If policy reload fails, the old policy remains while the hash is still advanced below, making the mismatch persistent and allowing an unresolved fail-open guard to be skipped. Please commit policy and registry atomically, or roll back and retry both together.

johntmyers · 2026-07-08T19:26:49Z

+        return Err("host pattern must not contain whitespace".to_string());
+    }
+
+    let pattern = glob::Pattern::new(&pattern.to_ascii_lowercase())


[P1] glob::Pattern allows * to cross dots, but the documented network endpoint semantics use * for one DNS label and ** for multiple labels. This can make an exclusion such as *.example.com unexpectedly suppress middleware for deep.api.example.com. Please reuse the label-aware DNS matcher and add multi-label exclusion coverage.

johntmyers · 2026-07-08T19:26:49Z

+        if name.is_empty()
+            || matches!(
+                name.as_str(),
+                "authorization" | "cookie" | "host" | "content-length" | "transfer-encoding"


[P1] Proxy-Authorization is not filtered here. For non-CONNECT forward-proxy requests, remote middleware can receive Basic proxy credentials before a later forwarding step strips the header. Please remove proxy credential headers before constructing the middleware evaluation.

Fixed in f4c0206. Proxy-Authorization is now filtered before constructing the middleware request, with regression coverage for both origin and proxy credentials.

johntmyers · 2026-07-08T19:26:49Z

+  // Destination and HTTP request target.
+  HttpRequestTarget target = 6;
+  // HTTP request headers before OpenShell injects credentials.
+  map<string, string> headers = 7;


[P1] A map cannot represent repeated HTTP headers. The current parser keeps only the last value for middleware, while request rebuilding preserves all original values upstream. That creates a first-value-versus-last-value authorization differential. Please preserve repeated header values in the contract or reject ambiguous duplicates before evaluation.

johntmyers · 2026-07-08T19:26:49Z

+        .iter()
+        .filter(|entry| entry.is_resolved())
+        .map(openshell_supervisor_middleware::DescribedChainEntry::max_body_bytes)
+        .min()


[P2] Taking the smallest cap for the entire chain loses per-stage on_error behavior. A 2 KiB request with a 1 KiB fail-open stage followed by a 1 MiB fail-closed stage is denied before the second stage runs, even though only the first stage should be skipped. Please buffer for eligible later stages and apply capacity failures per stage.

johntmyers · 2026-07-08T19:26:49Z

-                    "request body credential rewrite buffers at most {MAX_REWRITE_BODY_BYTES} bytes"
-                ));
-            }
+        if body.len().saturating_add(chunk_size) > MAX_REWRITE_BODY_BYTES {


[P2] This fixed credential-rewrite ceiling also applies to middleware chunked-body collection. A binding advertising a 1 MiB limit still rejects a 300 KiB chunked payload here. Please separate these limits or pass the negotiated middleware limit through the collector.

Fixed in f4c0206. Chunked middleware collection now uses the negotiated middleware body limit instead of the 256 KiB credential-rewrite ceiling; a 300 KiB payload under a 1 MiB binding cap is covered.

johntmyers · 2026-07-08T19:26:49Z

+        serde_json::Value::Bool(value) => Kind::BoolValue(value),
+        serde_json::Value::Number(value) => Kind::NumberValue(
+            value
+                .as_f64()


[P2] as_f64 accepts integers that cannot be represented exactly. For example, 9007199254740993 is delivered as 9007199254740992, which can change tenant IDs or allow/deny thresholds. Please reject values that do not round-trip exactly or use a lossless config representation.

Fixed in f4c0206. Integer config values that cannot round-trip exactly through protobuf double storage are now rejected, with coverage around the 2^53 boundary.

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

drew

Some initial feedback. I'll take another pass as well.

drew · 2026-07-09T06:46:11Z

+// MiddlewareManifest describes one service and the bindings it exposes.
+message MiddlewareManifest {
+  // Middleware protocol version implemented by the service.
+  string api_version = 1;


this probably isn't needed since the proto itself is versioned

drew · 2026-07-09T06:46:11Z

+  // Human-readable service name used for diagnostics.
+  string name = 2;
+  // Service-defined version string used for diagnostics.
+  string service_version = 3;


what is service in this context?

drew · 2026-07-09T06:46:11Z

+// ValidateConfigRequest contains one policy configuration to validate.
+message ValidateConfigRequest {
+  // Middleware protocol version selected by OpenShell.
+  string api_version = 1;


i dont think this is necessary if we're using protos

drew · 2026-07-09T06:46:11Z

+// HttpRequestEvaluation contains one buffered HTTP request to evaluate.
+message HttpRequestEvaluation {
+  // Middleware protocol version selected by OpenShell.
+  string api_version = 1;


same as above, the proto already has a version

drew · 2026-07-09T06:46:11Z

+use miette::{Result, miette};
+
+/// Binding identifier for the built-in secret redaction middleware.
+pub const BUILTIN_SECRETS: &str = "openshell/secrets";


is there a way to decouple this so that we don't hardcode middleware into openshell-core? this winds up on the client and other places i think.

maybe a redaction middleware crate that is configured into the server and supervisor?

drew · 2026-07-09T06:46:11Z

+        .iter()
+        .filter(|entry| entry.is_resolved())
+        .map(openshell_supervisor_middleware::DescribedChainEntry::max_body_bytes)
+        .min()


(agent finding)

[P2] Do not cap pre-buffering at the smallest middleware limit

When a chain mixes middleware with different max_body_bytes values, using the minimum here makes the relay treat any body larger than the smallest later middleware as unbufferable before ChainRunner can run earlier middleware that can handle it. For example, a 100 KB request with a first 256 KB redactor and a later 4 KB guard will either be denied or passed through unprocessed depending on on_error, instead of running the redactor and then applying the guard's per-entry capacity handling.

drew · 2026-07-09T06:47:33Z

nit: consider calling this supervisor_middleware.proto.

elezar · 2026-07-09T12:17:24Z

+  // gRPC endpoint reachable from the sandbox supervisor.
+  string grpc_endpoint = 2;
+  // Operator-owned body limit applied to every binding exposed by the service.
+  uint64 max_body_bytes = 4;


Agent review: SupervisorMiddlewareService is introduced in this PR and doesn't exist on main, so there are no deployed supervisors to maintain wire compatibility with. Renumber max_body_bytes from 4 to 3 to close the gap before this ships — a future contributor will otherwise reuse field 3 and silently break compatibility with any supervisors running this version.

elezar · 2026-07-09T12:18:44Z

+    {
+        return false;
+    }
+    name.starts_with("x-openshell-middleware-")


Agent review: is_safe_append_header uses a default-deny allowlist (correct security posture), but the required x-openshell-middleware-* prefix is not documented in the proto comment on add_headers, in the operator docs, or in the error message. An operator whose service returns x-correlation-id will get unsafe_response_headers routed through on_error with no explanation of what went wrong. The error at the validate_header_mutations call site should name the required prefix, and the proto comment and extensibility docs should document this constraint explicitly.

elezar · 2026-07-09T12:18:58Z

@@ -903,6 +949,42 @@ where
        let _ = &eval_target;

        if allowed || config.enforcement == EnforcementMode::Audit {


Agent review: when enforcement == Audit, middleware still runs and a Denied result sends a hard 403 to the client. Audit mode is defined as log-but-not-block — a middleware deny should be recorded as an OCSF finding and the request allowed through, not surfaced to the client as a rejection. Either the MiddlewareApplyResult::Denied arm should be treated as an audit finding here (and at the equivalent sites at lines 462, 1226, and 1456), or this intentional override of audit semantics needs to be explicitly documented and covered by a test.

elezar · 2026-07-09T12:25:18Z

+use miette::{Result, miette};
+
+/// Binding identifier for the built-in secret redaction middleware.
+pub const BUILTIN_SECRETS: &str = "openshell/secrets";


Non-blocking / follow-up suggestion.

Built-ins are currently treated structurally differently from external services: they bypass connect_services, are hardwired into IN_PROCESS_SERVICE, and require special-case branches in ensure_policy_bindings_registered and policy validation (== BUILTIN_SECRETS). Adding a second built-in also requires touching match arms in two separate places across two crates (openshell-core::middleware::validate_builtin_config and builtins/mod.rs).

A trait-based approach would address this: a BuiltinMiddleware trait (covering binding_id, describe, validate_config, evaluate) with a factory function instantiate(id) -> Option<&'static dyn BuiltinMiddleware> as the single registration point. Built-ins would self-register through the same connect_services path as externals, seeding the binding ID dedup set naturally and removing the special-case branches. Adding a new built-in would then only require implementing the trait and adding one arm to instantiate.

Not a blocker for this PR, but worth tracking before a second built-in lands.

pimlock force-pushed the 1733-supervisor-middleware/pmlocek branch 2 times, most recently from 595191e to 97b750f Compare June 26, 2026 23:58

pimlock self-assigned this Jun 30, 2026

pimlock force-pushed the 1733-supervisor-middleware/pmlocek branch from 358906a to 1fbcdbc Compare June 30, 2026 19:48

This comment was marked as outdated.

Sign in to view

pimlock added 18 commits July 2, 2026 15:17

feat(middleware): add in-process supervisor middleware

6ef9490

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(supervisor-middleware): harden middleware relay handling

e4e6f8f

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(supervisor-middleware): default stored policy rule fields

2765055

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(supervisor-middleware): resolve rebase policy conflicts

e487f2c

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

feat(supervisor-middleware): implement phase one runtime

9f892ef

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(supervisor-middleware): harden selection and buffering

da486b2

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

feat(supervisor-middleware): support external services

bb08a33

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

refactor(supervisor-middleware): clarify service contract

762887a

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

docs(supervisor-middleware): refine preview warning

31662d8

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

docs(extensibility): add supervisor middleware guide

2304e2f

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(server): remove stale middleware import

97bef95

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(network): remove needless test struct updates

ab2daba

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(middleware): avoid enabling core telemetry

a820dd2

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

refactor(supervisor-middleware): simplify service endpoints

49424a0

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

fix(supervisor-middleware): harden policy enforcement

9d4b631

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

feat(ocsf): enrich middleware shorthand logs

2b7cf4e

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

pimlock force-pushed the 1733-supervisor-middleware/pmlocek branch from c4b0dcf to 2b7cf4e Compare July 2, 2026 22:18

pimlock requested review from a team and derekwaynecarr as code owners July 3, 2026 18:34

refactor(ocsf): decouple shorthand from middleware

6fdc5e3

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

pimlock removed the test:e2e Requires end-to-end coverage label Jul 3, 2026

pimlock added 4 commits July 3, 2026 12:08

refactor(supervisor-middleware): clarify runtime integration

d27085c

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

refactor(middleware): move shared contracts to core

d2a9791

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

refactor(policy): extract middleware serialization

64439ea

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

refactor(middleware): move host matching to core

3b58759

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

pimlock mentioned this pull request Jul 3, 2026

rfc-0009: supervisor middleware #1738

Open

4 tasks

pimlock added the test:e2e Requires end-to-end coverage label Jul 3, 2026

feat(middleware): align operations and ordering

3b6fc10

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

pimlock mentioned this pull request Jul 7, 2026

feat(examples): add supervisor middleware content guard #2169

Draft

6 tasks

TaylorMutch reviewed Jul 7, 2026

View reviewed changes

elezar reviewed Jul 8, 2026

View reviewed changes

drew reviewed Jul 8, 2026

View reviewed changes

johntmyers reviewed Jul 8, 2026

View reviewed changes

ericksoa mentioned this pull request Jul 8, 2026

feat(sandbox): make L7 inference proxy CHUNK_IDLE_TIMEOUT configurable per route #866

Open

2 tasks

johntmyers reviewed Jul 8, 2026

View reviewed changes

fix(middleware): tighten request handling

f4c0206

Signed-off-by: Piotr Mlocek <pmlocek@nvidia.com>

drew reviewed Jul 9, 2026

View reviewed changes

Comment thread proto/middleware.proto

drew Jul 9, 2026

Copy link
Copy Markdown

Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: consider calling this supervisor_middleware.proto.

elezar reviewed Jul 9, 2026

View reviewed changes

		// HttpRequestTarget describes the admitted HTTP destination and request target.
		message HttpRequestTarget {

		@@ -903,6 +949,42 @@ where
		let _ = &eval_target;

		if allowed \|\| config.enforcement == EnforcementMode::Audit {

Uh oh!

Conversation

pimlock commented Jun 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Related Issue

Changes

Testing

Checklist

Uh oh!

copy-pr-bot Bot commented Jun 26, 2026

Uh oh!

github-actions Bot commented Jun 30, 2026

Uh oh!

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

pimlock commented Jul 2, 2026

Uh oh!

github-actions Bot commented Jul 3, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

pimlock Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

elezar Jul 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

drew left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

johntmyers left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

pimlock commented Jun 26, 2026 •

edited

Loading

pimlock Jul 7, 2026 •

edited

Loading

elezar Jul 8, 2026 •

edited

Loading

drew left a comment •

edited

Loading

johntmyers left a comment •

edited

Loading

drew left a comment •

edited

Loading