fix: wire GH_AW_AIC and GH_AW_MAX_AI_CREDITS to Handle agent failure step by Copilot · Pull Request #38327 · github/gh-aw

Copilot · 2026-06-10T10:34:17Z

Handle agent failure received GH_AW_AI_CREDITS_RATE_LIMIT_ERROR=true but never the actual credit values, so failure issues showed "AI Credits Budget Exceeded" with no usage or limit data. The noop step already had GH_AW_AIC; the failure step did not.

Root cause

resolveAICreditsFailureState() falls back to process.env.GH_AW_AIC / process.env.GH_AW_MAX_AI_CREDITS, but neither was in the step's env: block. shouldReportAICreditsRateLimitError(true, "", "") conservatively returns true when data is absent — correct behavior, wrong inputs.

Changes

pkg/workflow/notify_comment.go — adds two env vars to agentFailureEnvVars:
- GH_AW_AIC: ${{ needs.agent.outputs.aic }} — credits consumed
- GH_AW_MAX_AI_CREDITS — compile-time literal when max-ai-credits is in frontmatter; ${{ vars.GH_AW_DEFAULT_MAX_AI_CREDITS || '1000' }} otherwise (mirrors the firewall config logic)
actions/setup/js/ai_credits_context.cjs — adds provenance logging to resolveAICreditsFailureState(), reporting the source of each resolved value (audit_log / agent_stdio / env(GH_AW_AIC) / none) to the step log
245 lock files recompiled

…step, add provenance logging - `Handle agent failure` step was missing `GH_AW_AIC` (credits used) and `GH_AW_MAX_AI_CREDITS` (configured limit), causing "AI Credits Budget Exceeded" issues to show no data - Wire both env vars in `notify_comment.go` agentFailureEnvVars: AIC from agent job output, max from compile-time literal (when set in frontmatter) or runtime vars expression - Add provenance logging to `resolveAICreditsFailureState()` so each value's source (audit_log / agent_stdio / env / none) is visible in the step log Fixes #38322 and #38324 Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>

Copilot

Pull request overview

This PR fixes missing AI Credits usage/limit data in the Handle agent failure step by wiring the relevant env vars through to the failure handler and adding step-log provenance to aid debugging when credit data is absent.

Changes:

Pass GH_AW_AIC and GH_AW_MAX_AI_CREDITS into the agent-failure handler env so failure issues can include credits used + configured limit.
Add provenance logging in resolveAICreditsFailureState() to show which source (audit log / agent stdio / env / none) provided AI credits values.
Regenerate compiled workflow .lock.yml files to reflect the new env wiring.

Show a summary per file

File	Description
pkg/workflow/notify_comment.go	Threads AI credits usage (`GH_AW_AIC`) and budget limit (`GH_AW_MAX_AI_CREDITS`) into the agent-failure handler step env.
actions/setup/js/ai_credits_context.cjs	Logs provenance for resolved AI credits values during failure handling.
.github/workflows/*.lock.yml	Regenerated compiled workflows to include the new env vars in `Handle agent failure`.

Copilot's findings

Tip

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

Comments suppressed due to low confidence (1)

actions/setup/js/ai_credits_context.cjs:315

max-ai-credits supports the sentinel value -1 (disable enforcement), but this resolver only parses positive numbers. If GH_AW_MAX_AI_CREDITS is "-1" (from frontmatter or vars.GH_AW_DEFAULT_MAX_AI_CREDITS), maxAICredits becomes empty and shouldReportAICreditsRateLimitError will still conservatively flag an AI-credits budget failure; the failure handler will also log the configured max as "(none)". Consider treating "-1" as an explicit "no limit" value and forcing aiCreditsRateLimitError to false in that case.

Files reviewed: 247/247 changed files
Comments generated: 0

github-actions · 2026-06-10T10:57:12Z

✅ Design Decision Gate 🏗️ completed the design decision gate check.

No ADR enforcement needed: PR #38327 does not have the implementation label (has_implementation_label=false) and adds only 13 lines in business logic directories, well below the 100-line threshold (requires_adr_by_default_volume=false).

github-actions · 2026-06-10T10:57:13Z

🧪 Test Quality Sentinel completed test quality analysis.

No test files were added or modified in this PR. Test Quality Sentinel skipped.

github-actions · 2026-06-10T10:57:13Z

✅ PR Code Quality Reviewer completed the code quality review.

github-actions

REQUEST_CHANGES — one correctness bug must be fixed before merge

The core wiring is correct: GH_AW_AIC and GH_AW_MAX_AI_CREDITS were genuinely missing from the failure handler env block, and the provenance logging adds useful diagnostic signal. Three issues need attention:

### Blocking: wrong comparator lets the `-1` (unlimited) sentinel produce broken YAML

MaxAICredits != 0 is true for -1, which is the valid spec-supported unlimited sentinel (max-ai-credits: -1). The literal branch then emits GH_AW_MAX_AI_CREDITS: "-1", which parsePositiveNumberString rejects — leaving maxAICredits empty in the failure comment and the provenance log saying source=none while the var is clearly set. Fix: > 0 instead of != 0, and drop the dead data != nil conjunct while there. See inline comment.

### Non-blocking: missing tests and unreadable ternary

The new GH_AW_AIC / GH_AW_MAX_AI_CREDITS env-var wiring has no test coverage. The two-branch branching logic is the most failure-prone part and is entirely unasserted. See inline comment.
The rawRateLimitSignalSource ternary on line 309 of ai_credits_context.cjs is 200+ characters in a function that uses clean if/else if blocks for every other signal. See inline comment.

🔎 Code quality review by PR Code Quality Reviewer · ⌖ 13.5 AIC

github-actions · 2026-06-10T11:10:01Z

 	agentFailureEnvVars = append(agentFailureEnvVars, fmt.Sprintf("          GH_AW_EFFECTIVE_TOKENS: ${{ needs.%s.outputs.effective_tokens || '' }}\n", mainJobName))
 	agentFailureEnvVars = append(agentFailureEnvVars, fmt.Sprintf("          GH_AW_AI_CREDITS_RATE_LIMIT_ERROR: ${{ needs.%s.outputs.ai_credits_rate_limit_error || 'false' }}\n", mainJobName))
+	agentFailureEnvVars = append(agentFailureEnvVars, fmt.Sprintf("          GH_AW_AIC: ${{ needs.%s.outputs.aic }}\n", mainJobName))
+	if data != nil && data.EngineConfig != nil && data.EngineConfig.MaxAICredits != 0 {


Wrong comparator != 0 will silently discard the -1 (unlimited) sentinel and mislead the failure handler. When max-ai-credits: -1 is set (explicitly supported by the spec to disable budget enforcement), MaxAICredits != 0 is true, so the literal branch fires and emits GH_AW_MAX_AI_CREDITS: "-1" in the YAML. On the JS side, parsePositiveNumberString("-1") returns "" (it requires value > 0), so maxAICredits resolves to empty string and the provenance log records source=none GH_AW_MAX_AI_CREDITS=-1 — a confusing contradiction. The condition should be > 0.

💡 Suggested fix

// Before if data != nil && data.EngineConfig != nil && data.EngineConfig.MaxAICredits != 0 { // After if data.EngineConfig != nil && data.EngineConfig.MaxAICredits > 0 {

(Note: the data != nil guard can be dropped — data is unconditionally dereferenced many lines earlier in this function, e.g. at data.SafeOutputs on line 33, so it cannot be nil here. The extra conjunct is dead code that misleads readers into believing a nil-data code path exists.)

github-actions · 2026-06-10T11:10:01Z

+    console.log(`[ai-credits] maxAICredits source=none GH_AW_MAX_AI_CREDITS=${process.env.GH_AW_MAX_AI_CREDITS || "(unset)"}`);
+  }
+
+  const rawRateLimitSignalSource = auditRateLimitError ? "audit_log" : stdioSignals.rateLimitError ? "agent_stdio" : process.env.GH_AW_AI_CREDITS_RATE_LIMIT_ERROR === "true" ? "env(GH_AW_AI_CREDITS_RATE_LIMIT_ERROR)" : "none";


200+ character single-line ternary is unreadable and cannot be diffed. This is immediately above two if/else if blocks that handle aiCredits and maxAICredits identically — the inconsistency makes it look like an oversight.

💡 Suggested fix

// Before (one opaque line) const rawRateLimitSignalSource = auditRateLimitError ? "audit_log" : stdioSignals.rateLimitError ? "agent_stdio" : process.env.GH_AW_AI_CREDITS_RATE_LIMIT_ERROR === "true" ? "env(GH_AW_AI_CREDITS_RATE_LIMIT_ERROR)" : "none"; // After — consistent with the two blocks above it let rawRateLimitSignalSource; if (auditRateLimitError) { rawRateLimitSignalSource = "audit_log"; } else if (stdioSignals.rateLimitError) { rawRateLimitSignalSource = "agent_stdio"; } else if (process.env.GH_AW_AI_CREDITS_RATE_LIMIT_ERROR === "true") { rawRateLimitSignalSource = "env(GH_AW_AI_CREDITS_RATE_LIMIT_ERROR)"; } else { rawRateLimitSignalSource = "none"; }

github-actions · 2026-06-10T11:10:01Z

 	agentFailureEnvVars = append(agentFailureEnvVars, fmt.Sprintf("          GH_AW_AI_CREDITS_RATE_LIMIT_ERROR: ${{ needs.%s.outputs.ai_credits_rate_limit_error || 'false' }}\n", mainJobName))
+	agentFailureEnvVars = append(agentFailureEnvVars, fmt.Sprintf("          GH_AW_AIC: ${{ needs.%s.outputs.aic }}\n", mainJobName))
+	if data != nil && data.EngineConfig != nil && data.EngineConfig.MaxAICredits != 0 {
+		agentFailureEnvVars = append(agentFailureEnvVars, fmt.Sprintf("          GH_AW_MAX_AI_CREDITS: %q\n", strconv.FormatInt(data.EngineConfig.MaxAICredits, 10)))


No test asserts GH_AW_AIC or GH_AW_MAX_AI_CREDITS appear in the conclusion job env block. The two-branch logic (MaxAICredits > 0 literal vs. expression fallback) is completely uncovered. A regression that drops either line, misspells the key, or inverts the branch condition would pass CI undetected.

💡 What to test

Two new test cases in notify_comment_test.go:

Literal path — set EngineConfig.MaxAICredits = 500, call buildConclusionJob, assert the output contains GH_AW_MAX_AI_CREDITS: "500" and GH_AW_AIC: ${{ needs.agent.outputs.aic }}.

Expression path — leave EngineConfig.MaxAICredits = 0 (or EngineConfig = nil), assert the output contains GH_AW_MAX_AI_CREDITS: ${{ vars.GH_AW_DEFAULT_MAX_AI_CREDITS.

Unlimited sentinel (once the != 0 → > 0 fix is applied) — set MaxAICredits = -1, assert it falls through to the expression path rather than emitting "-1".

Copilot AI assigned Copilot and pelikhan Jun 10, 2026

Copilot created this pull request from a session on behalf of pelikhan June 10, 2026 10:34 View session

Copilot AI requested a review from pelikhan June 10, 2026 10:34

Copilot finished work on behalf of pelikhan June 10, 2026 10:34

pelikhan marked this pull request as ready for review June 10, 2026 10:43

Copilot AI review requested due to automatic review settings June 10, 2026 10:43

Copilot started reviewing on behalf of pelikhan June 10, 2026 10:43 View session

Copilot AI reviewed Jun 10, 2026

View reviewed changes

github-actions Bot mentioned this pull request Jun 10, 2026

[aw] No-Op Runs #38257

Closed

pelikhan merged commit 2446604 into main Jun 10, 2026
104 of 107 checks passed

pelikhan deleted the copilot/investigate-ai-credits-issue branch June 10, 2026 11:04

github-actions Bot requested changes Jun 10, 2026

View reviewed changes

github-actions Bot mentioned this pull request Jun 10, 2026

[docs] Update documentation for features from 2026-06-10 #38351

Merged

3 tasks

Copilot AI mentioned this pull request Jun 10, 2026

Merge conclusion job JS steps into a single conclusion_runner.cjs orchestrator #38369

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix: wire GH_AW_AIC and GH_AW_MAX_AI_CREDITS to Handle agent failure step#38327

fix: wire GH_AW_AIC and GH_AW_MAX_AI_CREDITS to Handle agent failure step#38327
pelikhan merged 1 commit into
mainfrom
copilot/investigate-ai-credits-issue

Copilot AI commented Jun 10, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

github-actions Bot commented Jun 10, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jun 10, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jun 10, 2026 •

edited

Loading

Uh oh!

Uh oh!

github-actions Bot left a comment

Uh oh!

github-actions Bot Jun 10, 2026

Uh oh!

github-actions Bot Jun 10, 2026

Uh oh!

github-actions Bot Jun 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Copilot AI commented Jun 10, 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

Copilot's findings

Uh oh!

github-actions Bot commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jun 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

github-actions Bot left a comment

Choose a reason for hiding this comment

REQUEST_CHANGES — one correctness bug must be fixed before merge

Uh oh!

github-actions Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Jun 10, 2026 •

edited

Loading

github-actions Bot commented Jun 10, 2026 •

edited

Loading

github-actions Bot commented Jun 10, 2026 •

edited

Loading

github-actions Bot commented Jun 10, 2026 •

edited

Loading