Fix firewall token usage reporting to preserve raw counts and remove cache-rate transforms

[Copilot]

• Inspect branch, main, and merge requirements
• Run baseline build/tests before changes
• Merge origin/main into this branch and resolve conflicts if any
• Recompile generated artifacts and commit only required updates
• Run targeted validation and reply to PR comment

[pelikhan]

@copilot merge main and recompile.

[Copilot]

Pull request overview

This pull request updates gh aw audit (and the setup action’s step summary) to stop deriving cache-efficiency / transformed totals from firewall token logs, and instead report raw token bucket counts while keeping Effective Tokens (ET) as the only computed metric.

Changes:

• Removes cache-efficiency calculation from Go and JS token-usage parsing paths.
• Updates audit rendering to print explicit raw token buckets (input/output/cache read/cache write) and removes “Total” / “Cache hit” lines.
• Updates Go and JS tests to assert the new “raw buckets + ET-only computed” behavior.

Show a summary per file

File	Description
pkg/cli/token_usage.go	Removes cache-efficiency computation during token usage parsing while keeping ET computation.
pkg/cli/token_usage_test.go	Updates expectations to ensure cache-efficiency is no longer computed.
pkg/cli/audit_report_render_findings.go	Changes audit token usage output to render raw token buckets and removes cache hit reporting.
pkg/cli/audit_report_test.go	Adds coverage to ensure audit rendering no longer prints transformed totals/cache hit lines.
actions/setup/js/parse_mcp_gateway_log.cjs	Removes cacheEfficiency from the parsed summary model and markdown footer.
actions/setup/js/parse_mcp_gateway_log.test.cjs	Updates tests to assert cacheEfficiency is not computed or rendered.
pkg/workflow/data/action_pins.json	Pins github/gh-aw-actions/setup@v0.71.5.
pkg/actionpins/data/action_pins.json	Pins github/gh-aw-actions/setup@v0.71.5 (duplicate pin source).
.github/workflows/issue-arborist.lock.yml	Updates locked workflow content (including codex harness model flag invocation).

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)

pkg/cli/token_usage.go:176

• CacheEfficiency is no longer computed, but TokenUsageSummary still carries a non-omitempty cache_efficiency float field. That means any JSON marshaling of TokenUsageSummary will now report cache_efficiency: 0 even when there are cache reads, which is misleading (and looks like an actual 0% metric rather than “not reported”). Consider removing the field from the summary entirely (and any downstream diff/rendering that uses it), or changing it to an optional/pointer field with omitempty so it’s absent when not computed/reported.

	tokenUsageLog.Printf("Parsed %d entries: %d input, %d output, %d cache_read, %d cache_write, %d requests",
		lineNum, summary.TotalInputTokens, summary.TotalOutputTokens,
		summary.TotalCacheReadTokens, summary.TotalCacheWriteTokens, summary.TotalRequests)

	// Compute effective tokens using per-model multipliers (with optional custom overrides)
	populateEffectiveTokensWithCustomWeights(summary, customWeights)

• Files reviewed: 9/9 changed files
• Comments generated: 1

[Copilot]

@copilot merge main and recompile.

Done in commit 3256b81: merged origin/main into this branch and recompiled workflow artifacts.