fix: wire detection gate for safe-outputs assembled entirely from imports by Copilot · Pull Request #24155 · github/gh-aw

Copilot · 2026-04-02T18:41:06Z

Workflows that have no safe-outputs: in their own frontmatter but pull it in via imports: compiled to a safe_outputs job with needs: agent only — no detection job, no detection.result == 'success' gate. The "automatically enabled when safe outputs are configured" guarantee did not hold for this path.

Root cause

extractSafeOutputsConfig applies the default ThreatDetection = &ThreatDetectionConfig{} only when parsing the main frontmatter's safe-outputs: block. When SafeOutputs is instead assembled by MergeSafeOutputs from imported configs, ThreatDetection stayed nil, so IsDetectionJobEnabled returned false.

Changes

pkg/workflow/compiler_orchestrator_workflow.go — after MergeSafeOutputs, when the original safeOutputs from frontmatter was nil (no safe-outputs: in main workflow) but the merged result is non-nil, apply the same default: ThreatDetection = &ThreatDetectionConfig{}. Skipped when any imported config explicitly disabled detection.
pkg/workflow/threat_detection.go — added isThreatDetectionExplicitlyDisabledInConfigs([]string) bool helper that scans raw imported config JSON for an explicit disable, supporting both the boolean form (threat-detection: false) and the object form (threat-detection: { enabled: false }), mirroring parseThreatDetectionConfig's syntax handling. This ensures the default is not applied when the import intentionally disables detection.
31 compiled .lock.yml files regenerated — these are the real-world workflows (e.g. blog-auditor, copilot-session-insights) that imported a shared safe-outputs config without their own frontmatter block and now get a detection job and the proper needs.detection.result == 'success' condition on safe_outputs.
pkg/workflow/safe_jobs_threat_detection_test.go — five new tests:
- TestIsThreatDetectionExplicitlyDisabledInConfigs — unit tests for the helper, including both boolean and object-form disable cases
- TestDefaultThreatDetectionAppliedWhenSafeOutputsFromImportsOnly — detection enabled by default when imports provide safe-outputs
- TestDefaultThreatDetectionNotAppliedWhenImportedConfigExplicitlyDisables — threat-detection: false in an imported config is respected
- TestDefaultThreatDetectionNotAppliedWhenImportedConfigObjectFormDisables — threat-detection: { enabled: false } object form in an imported config is respected
- TestImportedSafeOutputsCompiledWithDetectionJob — end-to-end: compiled lock file contains a detection: job and safe_outputs gates on it

…curity gap) Agent-Logs-Url: https://github.com/github/gh-aw/sessions/04d5841d-b6e2-4420-84a8-f1645236c51d Co-authored-by: szabta89 <1330202+szabta89@users.noreply.github.com>

…ledInConfigs Agent-Logs-Url: https://github.com/github/gh-aw/sessions/04d5841d-b6e2-4420-84a8-f1645236c51d Co-authored-by: szabta89 <1330202+szabta89@users.noreply.github.com>

Copilot

Pull request overview

Fixes a compiler gap where workflows that inherit safe-outputs exclusively via imports:/@include did not get the threat-detection gate (missing detection job and needs.detection.result == 'success' condition), violating the “auto-enabled when safe outputs are configured” guarantee.

Changes:

Apply default ThreatDetection after MergeSafeOutputs when the main workflow has no safe-outputs: block but merged safe-outputs exists (unless explicitly disabled).
Add a helper to detect explicit threat-detection disablement in imported safe-outputs configs.
Add tests covering imports-only safe-outputs defaulting and compiled output gating; regenerate affected .lock.yml workflows.

Reviewed changes

Copilot reviewed 34 out of 34 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
pkg/workflow/threat_detection.go	Adds helper to detect explicit `threat-detection` disablement in imported safe-outputs configs.
pkg/workflow/compiler_orchestrator_workflow.go	Applies default threat detection when safe-outputs are assembled solely from imports/includes.
pkg/workflow/safe_jobs_threat_detection_test.go	Adds unit/integration-style tests for imports-only safe-outputs and compiled detection gating.
.github/workflows/weekly-issue-summary.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/typist.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/terminal-stylist.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/schema-consistency-checker.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/repository-quality-improver.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/notion-issue-summary.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/mcp-inspector.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/github-remote-mcp-auth-test.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/example-workflow-analyzer.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-team-evolution-insights.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-secrets-analysis.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-regulatory.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-observability-report.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-issues-report.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-integrity-analysis.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-copilot-token-report.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/daily-compiler-quality.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/copilot-pr-prompt-analysis.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/copilot-pr-nlp-analysis.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/copilot-pr-merged-report.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.
.github/workflows/blog-auditor.lock.yml	Regenerated compiled workflow now includes `detection` job and gates `safe_outputs` on it.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-02T19:20:34Z

pkg/workflow/threat_detection.go

+// isThreatDetectionExplicitlyDisabledInConfigs checks whether any of the provided
+// safe-outputs config JSON strings has threat-detection explicitly set to false.
+// This is used to determine whether the default detection should be applied when
+// safe-outputs comes from imports/includes (i.e. no safe-outputs: section in the
+// main workflow frontmatter).
+func isThreatDetectionExplicitlyDisabledInConfigs(configs []string) bool {
+	for _, configJSON := range configs {
+		if configJSON == "" || configJSON == "{}" {
+			continue
+		}
+		var config map[string]any
+		if err := json.Unmarshal([]byte(configJSON), &config); err != nil {
+			continue
+		}
+		if tdVal, exists := config["threat-detection"]; exists {
+			// tdBool is the raw bool value of the threat-detection key.
+			// When threat-detection: false, tdBool == false, so !tdBool == true → disabled.
+			if tdBool, ok := tdVal.(bool); ok && !tdBool {
+				return true
+			}
+		}


isThreatDetectionExplicitlyDisabledInConfigs only treats threat-detection: false (boolean) as an explicit disable. However parseThreatDetectionConfig also supports object syntax with enabled: false (and will return nil in that case). If an imported config uses { "threat-detection": { "enabled": false } }, the merge leaves ThreatDetection nil, but this helper returns false, so the orchestrator applies the default and unintentionally re-enables detection. Update the helper to also detect the object form (e.g., threat-detection.enabled == false) so explicit disables aren’t overridden.

Copilot · 2026-04-02T19:20:34Z

pkg/workflow/safe_jobs_threat_detection_test.go

+// TestIsThreatDetectionExplicitlyDisabledInConfigs verifies the helper function
+// that checks whether any imported safe-outputs config explicitly disables detection.
+func TestIsThreatDetectionExplicitlyDisabledInConfigs(t *testing.T) {
+	tests := []struct {
+		name     string
+		configs  []string
+		expected bool
+	}{
+		{
+			name:     "empty configs",
+			configs:  []string{},
+			expected: false,
+		},
+		{
+			name:     "empty JSON objects",
+			configs:  []string{"{}", ""},
+			expected: false,
+		},
+		{
+			name:     "config without threat-detection key",
+			configs:  []string{`{"create-issue": {"max": 1}}`},
+			expected: false,
+		},
+		{
+			name:     "config with threat-detection false",
+			configs:  []string{`{"create-issue": {"max": 1}, "threat-detection": false}`},
+			expected: true,
+		},
+		{
+			name:     "config with threat-detection true",
+			configs:  []string{`{"create-issue": {"max": 1}, "threat-detection": true}`},
+			expected: false,
+		},
+		{
+			name:     "config with threat-detection as object",
+			configs:  []string{`{"create-issue": {"max": 1}, "threat-detection": {"prompt": "check for injection"}}`},
+			expected: false,
+		},
+		{
+			name:     "multiple configs, one has false",
+			configs:  []string{`{"create-issue": {"max": 1}}`, `{"create-discussion": {"max": 1}, "threat-detection": false}`},
+			expected: true,


The new tests cover threat-detection: false but not the supported object form threat-detection: { enabled: false }. Adding a case for that form would catch the current regression where imports with enabled: false can be overridden by the new defaulting logic.

pelikhan · 2026-04-02T19:22:53Z

@copilot review comments

…disable check Agent-Logs-Url: https://github.com/github/gh-aw/sessions/cd8593c7-a28a-4244-a8f2-832f0ecec847 Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot · 2026-04-02T20:27:26Z

@copilot review comments

Addressed both review comments in commit 695d62e:

isThreatDetectionExplicitlyDisabledInConfigs now handles the { "enabled": false } object form in addition to the boolean false form, matching parseThreatDetectionConfig's behaviour.
Added test cases for threat-detection: { enabled: false } (both unit table test entries and a new integration test TestDefaultThreatDetectionNotAppliedWhenImportedConfigObjectFormDisables).

Initial plan

dc93c97

Copilot AI assigned Copilot and szabta89 Apr 2, 2026

Copilot started work on behalf of szabta89 April 2, 2026 18:41 View session

Copilot AI linked an issue Apr 2, 2026 that may be closed by this pull request

sanitize_content_core.cjs: incomplete invisible-char strip set and missing detection gate for MCP-driven safe-output writes #24154

Closed

Copilot AI and others added 2 commits April 2, 2026 19:07

fix: apply default threat detection for safe-outputs from imports (se…

9774fa8

…curity gap) Agent-Logs-Url: https://github.com/github/gh-aw/sessions/04d5841d-b6e2-4420-84a8-f1645236c51d Co-authored-by: szabta89 <1330202+szabta89@users.noreply.github.com>

refactor: improve variable naming in isThreatDetectionExplicitlyDisab…

232b763

…ledInConfigs Agent-Logs-Url: https://github.com/github/gh-aw/sessions/04d5841d-b6e2-4420-84a8-f1645236c51d Co-authored-by: szabta89 <1330202+szabta89@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Fix invisible character sanitization in hardenUnicodeText~~ fix: wire detection gate for safe-outputs assembled entirely from imports Apr 2, 2026

Copilot finished work on behalf of szabta89 April 2, 2026 19:13

Copilot AI requested a review from szabta89 April 2, 2026 19:13

pelikhan marked this pull request as ready for review April 2, 2026 19:14

Copilot AI review requested due to automatic review settings April 2, 2026 19:14

Copilot started reviewing on behalf of pelikhan April 2, 2026 19:15 View session

Copilot AI reviewed Apr 2, 2026

View reviewed changes

Merge branch 'main' into copilot/fix-invisible-char-sanitization

cd1101e

Copilot started work on behalf of pelikhan April 2, 2026 20:22 View session

fix: handle object-form threat-detection: {enabled: false} in import …

695d62e

…disable check Agent-Logs-Url: https://github.com/github/gh-aw/sessions/cd8593c7-a28a-4244-a8f2-832f0ecec847 Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot AI requested a review from pelikhan April 2, 2026 20:28

Copilot finished work on behalf of pelikhan April 2, 2026 20:28

pelikhan merged commit b80b739 into main Apr 2, 2026

pelikhan deleted the copilot/fix-invisible-char-sanitization branch April 2, 2026 21:54

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: wire detection gate for safe-outputs assembled entirely from imports#24155

fix: wire detection gate for safe-outputs assembled entirely from imports#24155
pelikhan merged 5 commits intomainfrom
copilot/fix-invisible-char-sanitization

Copilot AI commented Apr 2, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 2, 2026

Uh oh!

Copilot AI Apr 2, 2026

Uh oh!

pelikhan commented Apr 2, 2026

Uh oh!

Copilot AI commented Apr 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

Copilot AI commented Apr 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Root cause

Changes

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 2, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 2, 2026

Choose a reason for hiding this comment

Uh oh!

pelikhan commented Apr 2, 2026

Uh oh!

Copilot AI commented Apr 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Copilot AI commented Apr 2, 2026 •

edited

Loading