Include last 5 tool calls in tool-denials failure issues by Copilot · Pull Request #39122 · github/gh-aw

Copilot · 2026-06-13T16:50:39Z

Tool-denial failure issues currently show only the terminal denied request (guard.tool_denials_exceeded reason), which hides the lead-up to the guardrail trip. This change adds the preceding tool-call context so triage can quickly identify the permission mismatch pattern.

Scope
- Extend tool_denials_exceeded issue context with the last 5 tool.execution_start calls seen before the guard event.
Event parsing updates (actions/setup/js/handle_agent_failure.cjs)
- While scanning session events.jsonl, collect tool calls from tool.execution_start.
- Persist recentToolCalls: recentToolCalls.slice(-5) on each captured guard.tool_denials_exceeded event.
- Keep existing denied-reason normalization behavior unchanged.
Issue rendering updates (actions/setup/md/tool_denials_exceeded_context.md)
- Add a new collapsible section:
  - Last 5 tool calls
- Render calls as markdown list items for quick scanning in failure issues.
Test updates (actions/setup/js/handle_agent_failure.test.cjs)
- Verify loader captures recent tool calls from session events.
- Verify context rendering includes the new section and entries.
- Verify truncation behavior keeps only the latest five calls.

events.push({
  denialCount,
  threshold,
  reason: parsed.data.reason?.trim() ?? "",
  recentToolCalls: recentToolCalls.slice(-5),
});

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

Copilot

Pull request overview

This PR improves triage for guard.tool_denials_exceeded failures by capturing and rendering the last 5 tool calls that occurred before the guardrail trip, so failure issues show the lead-up context instead of only the final denied request.

Changes:

Extend loadToolDenialsExceededEvents() to collect recent tool.execution_start calls and attach recentToolCalls to each captured guard event.
Update the tool_denials_exceeded_context.md template and renderer to include a new “Last 5 tool calls” collapsible section.
Add/extend tests to validate capture, rendering, and truncation behavior.

Show a summary per file

File	Description
actions/setup/md/tool_denials_exceeded_context.md	Adds a new collapsible section placeholder for rendering the last 5 tool calls.
actions/setup/js/handle_agent_failure.cjs	Collects tool call context from `events.jsonl` and renders it into the tool-denials-exceeded issue body.
actions/setup/js/handle_agent_failure.test.cjs	Adds coverage for capturing tool calls, rendering the new section, and truncating to the last 5.
.github/workflows/test-quality-sentinel.lock.yml	Updates generated lockfile content (includes a user-visible message emoji change).

Copilot's findings

Tip

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

Files reviewed: 4/4 changed files
Comments generated: 4

@@ -1262,6 +1272,7 @@ function loadToolDenialsExceededEvents() {
            denialCount,
            threshold,
            reason: typeof parsed.data.reason === "string" ? parsed.data.reason.trim() : "",
+            recentToolCalls: recentToolCalls.slice(-5),


  // Normalize the reason for display: multi-line programs (e.g. Python 3 heredocs) are
  // collapsed to a single-line summary so the issue body renders cleanly.
  const normalizedReason = normalizeDeniedPermissionCommand(reason);
+  const recentToolCallsList =
+    Array.isArray(latestEvent.recentToolCalls) && latestEvent.recentToolCalls.length > 0


        env:
          GH_AW_WORKFLOW_NAME: "Test Quality Sentinel"
-          GH_AW_SAFE_OUTPUT_MESSAGES: "{\"footer\":\"\\u003e 🧪 *Test quality analysis by [{workflow_name}]({run_url})*{ai_credits_suffix}{history_link}\",\"runStarted\":\"🔬 [{workflow_name}]({run_url}) is analyzing test quality on this {event_type}...\",\"runSuccess\":\"🧪 [{workflow_name}]({run_url}) completed test quality analysis.\",\"runFailure\":\"❌ [{workflow_name}]({run_url}) {status} during test quality analysis.\"}"
+          GH_AW_SAFE_OUTPUT_MESSAGES: "{\"footer\":\"\\u003e 🧪 *Test quality analysis by [{workflow_name}]({run_url})*{ai_credits_suffix}{history_link}\",\"runStarted\":\"🔬 [{workflow_name}]({run_url}) is analyzing test quality on this {event_type}...\",\"runSuccess\":\"✅ [{workflow_name}]({run_url}) completed test quality analysis.\",\"runFailure\":\"❌ [{workflow_name}]({run_url}) {status} during test quality analysis.\"}"


pelikhan · 2026-06-13T17:46:08Z

@copilot run pr-finisher skill

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

Copilot · 2026-06-13T18:07:47Z

@copilot run pr-finisher skill

Done. Here's what was addressed in commit f1c0b4e:

recentToolCalls unbounded growth — buffer is now capped at 5 with shift() on overflow, so it never grows beyond the window we need.

Guard event timestamp capture — loadToolDenialsExceededEvents now persists timestamp (ISO string) on each pushed event.

Nondeterministic latest-event selection — buildToolDenialsExceededContext now uses reduce to pick the event with the lexicographically greatest ISO timestamp instead of events[events.length - 1], which was order-dependent when multiple session directories were present.

Unrelated lock file change — reverted the 🧪 → ✅ runSuccess emoji change in test-quality-sentinel.lock.yml.

Tests updated to assert the new timestamp field in loader output. make fmt, make lint, make test-unit all pass locally.

CI needs a human re-trigger (close/reopen or workflow_dispatch) since agent pushes don't trigger workflows.

github-actions · 2026-06-13T18:19:14Z

@copilot review all comments and address unresolved review feedback. Then re-trigger CI after the open threads are resolved.

Generated by 👨‍🍳 PR Sous Chef · 143.8 AIC · ⌖ 1.04 AIC · ⊞ 17.4K · ◷

Copilot AI and others added 4 commits June 13, 2026 16:34

Plan tool-denials issue update

19a8305

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

Capture recent tool calls for denial issues

6a5f5d6

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

Format tool-denial context changes

709ddc8

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

Optimize tool-denials context parsing

569c31b

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

Copilot AI assigned Copilot and pelikhan Jun 13, 2026

Copilot created this pull request from a session on behalf of pelikhan June 13, 2026 16:51 View session

Copilot finished work on behalf of pelikhan June 13, 2026 16:51

Copilot AI requested a review from pelikhan June 13, 2026 16:51

pelikhan approved these changes Jun 13, 2026

View reviewed changes

pelikhan marked this pull request as ready for review June 13, 2026 16:53

Copilot AI review requested due to automatic review settings June 13, 2026 16:53

Copilot started reviewing on behalf of pelikhan June 13, 2026 16:53 View session

Copilot AI reviewed Jun 13, 2026

View reviewed changes

Copilot started work on behalf of pelikhan June 13, 2026 17:46 View session

Address review feedback: cap buffer, add timestamp, sort by ISO time,…

9b7b206

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

Copilot finished work on behalf of pelikhan June 13, 2026 18:08

Copilot AI requested a review from pelikhan June 13, 2026 18:08

pelikhan merged commit 4bacc0a into main Jun 13, 2026
14 checks passed

pelikhan deleted the copilot/update-agentic-failure-issue branch June 13, 2026 18:27

This was referenced Jun 14, 2026

[docs] Update documentation for features from 2026-06-14 #39214

Merged

[blog] Weekly blog post – 2026-06-15 #39318

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Include last 5 tool calls in tool-denials failure issues#39122

Include last 5 tool calls in tool-denials failure issues#39122
pelikhan merged 5 commits into
mainfrom
copilot/update-agentic-failure-issue

Copilot AI commented Jun 13, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

pelikhan commented Jun 13, 2026

Uh oh!

Copilot AI commented Jun 13, 2026

Uh oh!

github-actions Bot commented Jun 13, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Copilot AI commented Jun 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Copilot's findings

Uh oh!

Uh oh!

pelikhan commented Jun 13, 2026

Uh oh!

Copilot AI commented Jun 13, 2026

Uh oh!

github-actions Bot commented Jun 13, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Jun 13, 2026 •

edited

Loading