Isolate threat detection provider settings across engines by Copilot · Pull Request #66533 · github/gh-aw

Copilot · 2026-10-07T11:39:15Z

Threat detection inherited the main engine’s provider environment even when using a different engine, leaking credentials and causing firewall health checks to fail before analysis. Missing verdicts could remain unnoticed while safe outputs were published.

Provider isolation: Inherit engine.env, API targets, and custom model aliases only when engines match. Preserve explicit detector overrides and built-in aliases; use the detector’s default model when an inherited alias is discarded.
Consistent configuration: Apply isolation to execution, firewall domain propagation, job dependencies, and authentication checks.
Missing-verdict visibility: Warn in the conclusion job on engine failures, unparseable results, or absent conclusions—even after an output comment is posted. Preserve duplicate-comment suppression and continue-on-error behavior.

This split-engine configuration no longer requires empty provider-env overrides:

engine:
  id: claude
  env:
    ANTHROPIC_BASE_URL: https://openrouter.ai/api
    ANTHROPIC_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
safe-outputs:
  threat-detection:
    engine:
      id: copilot
      model: sonnet

Fixes Threat detection inherits the main engine's provider env when it runs on a different engine #66203

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

pelikhan · 2026-10-07T11:59:46Z

@copilot automatically match the main agent provider if the agentic engine do not match (OpenAI/anthropic/copilot...).
A user of the pi engine should NOT have to reconfigure the threat detection job. Good default must be in place.

…model aliases Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot · 2026-10-07T12:31:03Z

@copilot automatically match the main agent provider if the agentic engine do not match (OpenAI/anthropic/copilot...).
A user of the pi engine should NOT have to reconfigure the threat detection job. ...

Implemented in 5568fd15: default detection now follows the main engine’s effective provider, and matching provider settings are inherited across engine IDs. Pi workflows automatically select the corresponding built-in detector; explicit detection-engine configuration still takes precedence.

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot

Copilot review overview

🟡 Changes recommended

Same-provider credentials and API targets still cross differing engine identities, contradicting the documented security contract.

Review effort: Balanced
Findings: 1

Open (1)

Restrict threat detection credential inheritance to matching engine IDs · New

What changed in this PR

This PR isolates threat-detection provider configuration and surfaces missing security verdicts.

Changes:

Restricts provider configuration inheritance and updates firewall/auth resolution.
Warns when detection produces no verdict.
Documents behavior and regenerates compiled workflows.

File	Description
`pkg/workflow/compiler_validators.go`	Uses effective detector environment for auth validation.
`pkg/workflow/evals_steps.go`	Applies engine-aware firewall-domain resolution.
`pkg/workflow/threat_detection_firewall_test.go`	Updates firewall helper tests.
`docs/src/content/docs/reference/threat-detection.md`	Documents isolation and missing-verdict warnings.
295 `.github/workflows/*.lock.yml` files	Propagate the generated conclusion warning.

Copilot · 2026-10-07T14:15:54Z


+When the resolved detection engine matches the main engine, detection inherits `engine.env`, `api-target`, and model aliases. Detection-specific environment values take precedence, including empty values.
+
+When the engines differ, detection does not inherit the main engine's environment, API target, or custom model aliases. This prevents provider URLs, API keys, and custom headers from reaching an unrelated engine. Configure any required provider settings under `safe-outputs.threat-detection.engine.env`; built-in model aliases remain available.


Fixed in 9329e85f: main-engine environment and API-target settings now inherit only when the resolved engine IDs match. Regression coverage confirms Pi/OpenAI settings do not reach Codex detection while provider-based detector selection and explicit detector overrides remain supported.

gh-aw-bot · 2026-10-07T13:53:58Z

@copilot address the following outstanding work in one pass:

Update this branch with the latest main using make merge-main, resolving any conflicts and preserving the intended changes.
Review (docs/src/content/docs/reference/threat-detection.md:249): This documented isolation is not what the implementation enforces. mergeThreatDetectionEngineEnv still copies provider-prefixed credentials and headers when different engine IDs resolve to the same provider, and both detector paths also inherit APITarget through sameThreatDetectionProvider. For example, a Pi/OpenAI main engine running Codex detection receives the main OPENAI_API_KEY by design in the new test. That leaves credentials crossing the engine boundary, contrary to this line, the PR description, and issue Threat detection inherits the main engine's provider env when it runs on a different engine #66203. Please restrict environment/API-target inheritance to matching engine IDs (while keeping detector-specific overrides), or revise the stated security contract if cross-engine provider inheritance is actually intended. - Isolate threat detection provider settings across engines #66533 (comment)

Push the necessary fixes, reply to each listed review thread and resolve it when addressed. Ignore feedback already answered or resolved. Use the pr-finisher skill and stop when only human review or CI remains; do not trigger CI.

Sous-chef head: 9fd53c8
Sous-chef work: 2ee654c89d8eb9fd5d652a665b5da0464b4005b587887582215b0555dfe6f95b
Sous-chef state: 75156fe1b8f656934acf11c6ea9250ed4e1d4776574b41e284ffa953c8262492

Generated by 👨‍🍳 PR Sous Chef · pi · haiku45 · 7.92 AIC · ⌖ 6.03 AIC · ⊞ 1K · ◷
_{Comment /souschef to run again}

…tection-integration Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>

Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>

github-actions · 2026-10-07T20:32:09Z

🎉 This pull request is included in a new release.

Release: v0.91.5

Initial plan

8756039

Copilot AI assigned Copilot and pelikhan Oct 7, 2026

Copilot AI linked an issue Oct 7, 2026 that may be closed by this pull request

Threat detection inherits the main engine's provider env when it runs on a different engine #66203

Closed

Copilot started work on behalf of pelikhan October 7, 2026 11:39 View session

Isolate threat detection provider settings across engine identities

8c9f4a7

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot started work on behalf of pelikhan October 7, 2026 12:01 View session

Keep missing-verdict warnings visible and avoid unresolved detection …

35d1d77

…model aliases Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Fix threat detection inheritance from main engine provider env~~ Isolate threat detection provider settings across engines Oct 7, 2026

Copilot finished work on behalf of pelikhan October 7, 2026 12:06

Copilot AI requested a review from pelikhan October 7, 2026 12:06

Match threat detection to main provider

5568fd1

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Apply remaining changes

9fd53c8

Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot finished work on behalf of pelikhan October 7, 2026 12:32

pelikhan marked this pull request as ready for review October 7, 2026 13:27

Copilot AI balanced review requested due to automatic review settings October 7, 2026 13:27

Copilot started reviewing on behalf of pelikhan October 7, 2026 13:28 View session

Copilot AI reviewed Oct 7, 2026

View reviewed changes

github-actions Bot mentioned this pull request Oct 7, 2026

[pr-sous-chef] 🍳 PR Sous Chef Summary: 4 PRs nudged toward completion #66387

Closed

Copilot started work on behalf of gh-aw-bot October 7, 2026 13:54 View session

Copilot AI and others added 2 commits October 7, 2026 14:00

Merge remote-tracking branch 'origin/main' into copilot/fix-threat-de…

2074ca3

…tection-integration Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>

Restrict threat detection inheritance to engine IDs

9329e85

Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>

Copilot finished work on behalf of gh-aw-bot October 7, 2026 14:20

Copilot AI requested a review from gh-aw-bot October 7, 2026 14:20

pelikhan merged commit 9bb87b2 into main Oct 7, 2026
36 checks passed

pelikhan deleted the copilot/fix-threat-detection-integration branch October 7, 2026 15:49

github-actions Bot mentioned this pull request Oct 8, 2026

[gh-aw-parity] gh-aw threat-detection changes to review - 2026-10-08 github/gh-aw-threat-detection#1227

Open

Navigation Menu

Repository navigation

Isolate threat detection provider settings across engines - #66533

Isolate threat detection provider settings across engines#66533
pelikhan merged 7 commits into
mainfrom
copilot/fix-threat-detection-integration

Copilot AI commented Oct 7, 2026 •

edited

Loading

Uh oh!

pelikhan commented Oct 7, 2026

Uh oh!

Copilot AI commented Oct 7, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Oct 7, 2026

Uh oh!

gh-aw-bot commented Oct 7, 2026

Uh oh!

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants


		When the resolved detection engine matches the main engine, detection inherits `engine.env`, `api-target`, and model aliases. Detection-specific environment values take precedence, including empty values.

		When the engines differ, detection does not inherit the main engine's environment, API target, or custom model aliases. This prevents provider URLs, API keys, and custom headers from reaching an unrelated engine. Configure any required provider settings under `safe-outputs.threat-detection.engine.env`; built-in model aliases remain available.

Conversation

Copilot AI commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pelikhan commented Oct 7, 2026

Uh oh!

Copilot AI commented Oct 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Copilot AI Oct 7, 2026

Choose a reason for hiding this comment

Uh oh!

gh-aw-bot commented Oct 7, 2026

Uh oh!

Uh oh!

github-actions Bot commented Oct 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Copilot AI commented Oct 7, 2026 •

edited

Loading