feat: add ext refac and sandbox

.compozy/tasks/ext-parity/reviews-001/_meta.md

@@ -0,0 +1,11 @@
+---
+provider: coderabbit
+pr: "25"
+round: 1
+created_at: 2026-04-16T16:42:11.902095Z
+---
+
+## Summary
+- Total: 46
+- Resolved: 46
+- Unresolved: 0

.compozy/tasks/ext-parity/reviews-001/issue_001.md

@@ -0,0 +1,202 @@
+---
+status: resolved
+file: extensions/bridges/discord/provider.go
+line: 2351
+author: coderabbitai[bot]
+provider_ref: thread:PRRT_kwDOR5y4QM57fTPp,comment:PRRC_kwDOR5y4QM64dqFo
+---
+
+# Issue 001: _⚠️ Potential issue_ | _🟠 Major_
+## Review Comment
+
+_⚠️ Potential issue_ | _🟠 Major_
+
+<details>
+<summary>🧩 Analysis chain</summary>
+
+🏁 Script executed:
+
+```shell
+#!/bin/bash
+set -euo pipefail
+
+echo "Check where not-initialized RPC errors are produced:"
+rg -nP --type=go 'NewRPCError\(\s*(rpcCodeNotInitialized|-32003)\b'
+
+echo
+echo "Check for string-based matching of initialization errors:"
+rg -nP --type=go 'Not initialized'
+```
+
+Repository: compozy/agh
+
+Length of output: 5929
+
+---
+
+🏁 Script executed:
+
+```shell
+# Read the isNotInitializedRPCError function from discord/provider.go
+sed -n '2320,2360p' extensions/bridges/discord/provider.go
+```
+
+Repository: compozy/agh
+
+Length of output: 969
+
+---
+
+🏁 Script executed:
+
+```shell
+# Check if there are tests for isNotInitializedRPCError in discord
+rg -n 'isNotInitializedRPCError' extensions/bridges/discord/
+```
+
+Repository: compozy/agh
+
+Length of output: 668
+
+---
+
+🏁 Script executed:
+
+```shell
+# Verify if the function signature and behavior are tested
+rg -B5 -A10 'isNotInitializedRPCError' extensions/bridges/discord/provider_test.go | head -50
+```
+
+Repository: compozy/agh
+
+Length of output: 718
+
+---
+
+🏁 Script executed:
+
+```shell
+# Find all test usages of isNotInitializedRPCError in discord tests
+rg -B2 -A2 'isNotInitializedRPCError' extensions/bridges/discord/provider_test.go
+```
+
+Repository: compozy/agh
+
+Length of output: 379
+
+---
+
+🏁 Script executed:
+
+```shell
+# Check if there are any references to rpcCodeErr type definition in discord tests
+rg -B5 'type rpcCodeErr' extensions/bridges/discord/provider_test.go
+```
+
+Repository: compozy/agh
+
+Length of output: 131
+
+---
+
+🏁 Script executed:
+
+```shell
+# Get the full rpcCodeErr type definition including its Code() method
+rg -A5 'type rpcCodeErr struct' extensions/bridges/discord/provider_test.go
+```
+
+Repository: compozy/agh
+
+Length of output: 206
+
+---
+
+
+
+</details>
+
+**Remove string-based RPC error matching fallback in `isNotInitializedRPCError`.**
+
+The function currently falls back to string comparison (`strings.EqualFold(rpcErr.Message, "Not initialized")`), which is brittle and violates the coding guideline: "never compare error strings." Production code consistently uses the error code (`rpcCodeNotInitialized`), and multiple bridges use different message variants ("Not initialized", "not ready", etc.), making the message fallback unreliable. Keep matching strictly code-based via the `Code()` interface check.
+
+<details>
+<summary>🔧 Proposed fix</summary>
+
+```diff
+ func isNotInitializedRPCError(err error) bool {
+ 	if err == nil {
+ 		return false
+ 	}
+ 	type rpcCodeError interface {
+ 		Code() int
+ 	}
+ 	var codeErr rpcCodeError
+ 	if errors.As(err, &codeErr) && codeErr.Code() == rpcCodeNotInitialized {
+ 		return true
+ 	}
+ 	var rpcErr *subprocess.RPCError
+-	if !errors.As(err, &rpcErr) {
+-		return false
+-	}
+-	return rpcErr.Code == rpcCodeNotInitialized ||
+-		strings.EqualFold(strings.TrimSpace(rpcErr.Message), "Not initialized")
++	return errors.As(err, &rpcErr) && rpcErr.Code == rpcCodeNotInitialized
+ }
+```
+</details>
+
+<!-- suggestion_start -->
+
+<details>
+<summary>📝 Committable suggestion</summary>
+
+> ‼️ **IMPORTANT**
+> Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
+
+```suggestion
+	if err == nil {
+		return false
+	}
+	type rpcCodeError interface {
+		Code() int
+	}
+	var codeErr rpcCodeError
+	if errors.As(err, &codeErr) && codeErr.Code() == rpcCodeNotInitialized {
+		return true
+	}
+	var rpcErr *subprocess.RPCError
+	return errors.As(err, &rpcErr) && rpcErr.Code == rpcCodeNotInitialized
+```
+
+</details>
+
+<!-- suggestion_end -->
+
+<details>
+<summary>🤖 Prompt for AI Agents</summary>
+
+```
+Verify each finding against the current code and only fix it if needed.
+
+In `@extensions/bridges/discord/provider.go` around lines 2336 - 2351, The
+function isNotInitializedRPCError currently falls back to brittle string
+matching on subprocess.RPCError.Message; remove that string-based check and
+return only based on the code-based checks: keep the errors.As check for the
+Code() interface (rpcCodeError) and the check against rpcCodeNotInitialized, and
+if errors.As into *subprocess.RPCError, only use rpcErr.Code ==
+rpcCodeNotInitialized to decide true/false; delete the strings.EqualFold(...
+"Not initialized") fallback and ensure the function returns false when no code
+match is found (no other error-string comparisons).
+```
+
+</details>
+
+<!-- fingerprinting:phantom:poseidon:hawk -->
+
+<!-- This is an auto-generated comment by CodeRabbit -->
+
+## Triage
+
+- Decision: `VALID`
+- Notes: `isNotInitializedRPCError` still falls back to matching `subprocess.RPCError.Message == "Not initialized"`, which is brittle and inconsistent with the code-based path already present in the same helper. The fix is to remove the string fallback and tighten the discord tests so only RPC code-based matches succeed.

.compozy/tasks/ext-parity/reviews-001/issue_002.md

@@ -0,0 +1,21 @@
+---
+status: resolved
+file: extensions/bridges/github/provider_test.go
+line: 1050
+severity: nitpick
+author: coderabbitai[bot]
+provider_ref: review:4122628916,nitpick_hash:e4870173171b
+review_hash: e4870173171b
+source_review_id: "4122628916"
+source_review_submitted_at: "2026-04-16T16:31:31Z"
+---
+
+# Issue 002: Test name no longer matches behavior
+## Review Comment
+
+The test is still named `TestGitHubProviderReconcileRejectsSharedWebhookPaths`, but it now asserts shared paths are accepted. Please rename it to reflect the new contract (e.g., `...AllowsSharedWebhookPaths`) to avoid future misreads.
+
+## Triage
+
+- Decision: `VALID`
+- Notes: The test name still says shared webhook paths are rejected, but the assertions explicitly require both configs to be accepted. That mismatch is real and will be corrected while fixing the shared-path signature-routing bug covered by issue 003.

.compozy/tasks/ext-parity/reviews-001/issue_003.md

@@ -0,0 +1,108 @@
+---
+status: resolved
+file: extensions/bridges/github/provider_test.go
+line: 1054
+author: coderabbitai[bot]
+provider_ref: thread:PRRT_kwDOR5y4QM57fTP2,comment:PRRC_kwDOR5y4QM64dqF6
+---
+
+# Issue 003: _⚠️ Potential issue_ | _🟠 Major_
+## Review Comment
+
+_⚠️ Potential issue_ | _🟠 Major_
+
+<details>
+<summary>🧩 Analysis chain</summary>
+
+🏁 Script executed:
+
+```shell
+#!/bin/bash
+set -euo pipefail
+
+echo "Inspect webhook signature + routing flow in provider implementation..."
+rg -n -C6 'func verifyGitHubWebhookSignature|func selectGitHubIssueConfig|func \(.*\) handleWebhookRequest|webhookSecret|X-Hub-Signature-256' extensions/bridges/github/provider.go
+
+echo
+echo "Look for any-candidate signature acceptance loops:"
+rg -n -C5 'for _, .* := range candidates|ConstantTimeCompare|hmac|sha256' extensions/bridges/github/provider.go
+```
+
+Repository: compozy/agh
+
+Length of output: 5390
+
+---
+
+🏁 Script executed:
+
+```shell
+#!/bin/bash
+set -euo pipefail
+
+echo "=== Examine handleWebhookRequest and how signature verification is used ==="
+rg -n -C10 'handleWebhookRequest' extensions/bridges/github/provider.go | head -80
+
+echo
+echo "=== Look for where verifyGitHubWebhookSignature is called ==="
+rg -n -B5 -A10 'verifyGitHubWebhookSignature' extensions/bridges/github/provider.go | grep -A15 'verifyGitHubWebhookSignature('
+
+echo
+echo "=== Examine the test context around line 1050 ==="
+sed -n '1010,1080p' extensions/bridges/github/provider_test.go
+```
+
+Repository: compozy/agh
+
+Length of output: 4636
+
+---
+
+
+
+</details>
+
+**Fix test name-behavior mismatch and address signature-validation binding gap**
+
+The test name `TestGitHubProviderReconcileRejectsSharedWebhookPaths` contradicts the assertions (line 1050-1054), which now expect `configError == nil`—accepting shared paths.
+
+More critically, the current signature validation (line 1648 in `verifyGitHubWebhookSignature`) accepts ANY candidate's secret, but routing (line 1562 in `selectGitHubIssueConfig`) selects by repository name from the webhook payload. This means a payload signed with Instance A's secret can be routed to Instance B if the `repository.full_name` field matches Instance B—Instance B's secret is never verified against the signature.
+
+Either:
+1. Rename the test to reflect acceptance of shared paths, OR
+2. Restore rejection of shared webhook paths, OR  
+3. Change signature verification to re-validate against the selected instance's secret after routing
+
+<details>
+<summary>🤖 Prompt for AI Agents</summary>
+
+```
+Verify each finding against the current code and only fix it if needed.
+
+In `@extensions/bridges/github/provider_test.go` around lines 1050 - 1054, The
+test name TestGitHubProviderReconcileRejectsSharedWebhookPaths is inconsistent
+with assertions that expect configs[i].configError == nil and there is a real
+security gap: verifyGitHubWebhookSignature currently accepts any candidate
+secret but selectGitHubIssueConfig routes by repository.full_name, so an event
+signed with Instance A's secret can be delivered to Instance B without
+re-checking its secret. Fix by (preferably) updating the webhook handling flow:
+after selectGitHubIssueConfig selects the target config (by
+repository.full_name), re-run signature verification against that selected
+config's secret (use the same verification logic in verifyGitHubWebhookSignature
+but with the selectedConfig.Secret) before accepting/routing the payload; also
+update tests (either rename TestGitHubProviderReconcileRejectsSharedWebhookPaths
+to reflect acceptance, or change tests to assert rejection if you restore
+shared-path rejection) so they match the new behavior and add a unit test that
+ensures a payload signed with A's secret is rejected for B.
+```
+
+</details>
+
+<!-- fingerprinting:phantom:poseidon:hawk -->
+
+<!-- This is an auto-generated comment by CodeRabbit -->
+
+## Triage
+
+- Decision: `VALID`
+- Notes: The production routing gap is real. `serveWebhookHTTP` verifies the signature against any config sharing the path, then `selectGitHubIssueConfig`/`selectGitHubReviewConfig` routes by payload repository name, so a payload signed with one instance secret can be routed to another instance on the same webhook path. Fix requires a minimal production change in `extensions/bridges/github/provider.go` to re-verify the selected config's secret after routing, plus test updates in `provider_test.go`.

.compozy/tasks/ext-parity/reviews-001/issue_004.md

@@ -0,0 +1,21 @@
+---
+status: resolved
+file: extensions/bridges/teams/provider_test.go
+line: 1255
+severity: nitpick
+author: coderabbitai[bot]
+provider_ref: review:4122628916,nitpick_hash:bcca49a3ba22
+review_hash: bcca49a3ba22
+source_review_id: "4122628916"
+source_review_submitted_at: "2026-04-16T16:31:31Z"
+---
+
+# Issue 004: Wait for ready state in the polling predicate to reduce flakiness.
+## Review Comment
+
+This currently waits for “any state” and asserts readiness afterward. If an intermediate non-ready state is written first, the test can fail intermittently.
+
+## Triage
+
+- Decision: `VALID`
+- Notes: The current predicate returns as soon as any state marker exists, but the assertion checks readiness afterward. If the file first contains a non-ready marker and later appends a ready marker, this test can observe the intermediate state and fail. The fix is to wait for a ready marker in the predicate.

.compozy/tasks/ext-parity/reviews-001/issue_005.md

@@ -0,0 +1,21 @@
+---
+status: resolved
+file: go.mod
+line: 77
+severity: nitpick
+author: coderabbitai[bot]
+provider_ref: review:4122628916,nitpick_hash:c42d9123128f
+review_hash: c42d9123128f
+source_review_id: "4122628916"
+source_review_submitted_at: "2026-04-16T16:31:31Z"
+---
+
+# Issue 005: Consider updating the daytona SDK if gorilla/websocket pseudo-version drift is a concern.
+## Review Comment
+
+The `gorilla/websocket` pseudo-version is an indirect dependency pulled transitively through `github.com/daytonaio/daytona/libs/sdk-go/pkg/daytona` with no direct imports in this codebase. The version constraint is controlled by the daytona SDK, not by this project. If supply-chain risk from the pseudo-version is a concern, you would need to update the daytona SDK version or explicitly pin `gorilla/websocket` at the expense of potentially conflicting with daytona's requirements.
+
+## Triage
+
+- Decision: `INVALID`
+- Notes: This is dependency-management advice, not a defect in the scoped change. `github.com/gorilla/websocket` is only an indirect dependency here and the review comment does not identify a concrete incompatibility, failing build, or exploitable issue caused by the current graph. Updating the upstream Daytona SDK is outside this remediation batch.