[copilot-agent-analysis] Daily Copilot Agent Analysis - 2026-06-13

[github-actions[bot]]

🤖 Copilot Agent PR Analysis - 2026-06-13

Analysis Period: Last 24 hours
Total PRs (agent_prs_total): 49 | Merged (agent_prs_merged): 39 (80%) | Avg Duration: ~110 min | Avg Human Comments: 0.86

Performance Metrics (by calendar day)

Date	PRs	Merged	Closed	Open	Success Rate	Avg Duration
2026-06-13	37	30	3	4	81%	76 min
2026-06-12	31	26	4	1	84%	274 min
2026-06-11	49	32	17	0	65%	124 min

Trend: 📈 Success rate improved from 65% (06-11) to 81% (06-13), +16 pts — 06-11 carried an unusually high closure count (17), now normalized to 3–4/day.

Agent Task Texts (49 PRs, last 24h)

PR #	Status	Task Text (first 90 chars)
#39130	Open	actions/cache is branch-scoped: caches written by the conclusion job on one PR branch ar
#39129	Merged	This change adds compiler guidance for Copilot-engine workflows to prefer GitHub Actions t
#39128	Merged	max-ai-credits: 5M appeared in multiple docs examples and created confusing guidance. Th
#39123	Merged	The daily max AI-credit guardrail was incorrectly applying to runs where users intentional
#39122	Merged	Tool-denial failure issues currently show only the terminal denied request (`guard.tool_de
#39119	Merged	Persona exploration showed strong baseline guidance but identified three doc gaps: compact
#39118	Open	The 1 MB default for max-patch-size on create-pull-request and `push-to-pull-request-b
#39102	Merged	gh aw forecast was sampling only successful runs, which undercounted AI credit usage for
#39101	Merged	The forecast report was undercounting workflows that spend AI credits in threat-detection,
#39100	Open	This changes safe-outputs from a separately started HTTP sidecar into a first-class MCP se
#39099	Merged	test-quality-sentinel used the same 🧪 emoji for both workflow attribution (footer) and
#39098	Merged	Appendix A showed token-class inputs and a final cost_usd/aic result, but omitted the
#39097	Closed	Every compilation triggered a full yaml.Unmarshal on ~80KB of generated YAML despite sch
#39096	Merged	The release workflow’s Defender stage failed on transient MpCmdRun -SignatureUpdate erro
#39089	Open	The Daily Issues Report workflow failed because the Copilot engine exited before producing
#39086	Merged	The publish-safe-outputs-node.yml workflow was blocking image publishes for 7 days after
#39085	Merged	The four run: scripts in publish-safe-outputs-node.yml produced zero visible output —
#39084	Merged	Cache entries were stored as {run_id, aic} with no timestamp, making age-based pruning i
#39081	Merged	This adds a dedicated gh-aw-node container image built on node:lts-alpine with git,
#39080	Merged	safe-outputs.dispatch-workflow already enforced cross-repo allowlists at runtime, but th
#39069	Merged	Agent failure issues caused by guardrail limits (max-ai-credits, max-daily-ai-credits,
#39068	Merged	Adds an experiment campaign to architecture-guardian testing whether inlining violation
#39066	Merged	When the daily AIC guardrail blocks a run, the agent never executes and no usage files are
#39064	Merged	Static analysis was stable except for persistent high-volume zizmor template-injection
#39061	Merged	The mattpocock-skills-reviewer workflow was blocked after consuming 5.4K AI Credits agai
#39058	Merged	The daily-cli-tools-tester workflow had no explicit max-ai-credits setting, defaulting
#39057	Merged	The Copilot CLI Deep Research workflow was failing early because repeated shell permission
#39056	Merged	The Daily Workflow Updater (run [#27461126080](https://github.com/github/gh-aw/actions/run
#39054	Merged	Three production *http.Client.Get calls omitted context.Context, making in-flight requ
#39053	Merged	Several trigger-specific helpers in frontmatter_extraction_yaml.go repeated the same `fr
#39052	Merged	The Failure Investigator prefetch could return an empty failed_run_ids set even when fai
#39051	Merged	add-comment previously requested discussions: write unless explicitly disabled, which
#39012	Merged	Regenerates all .lock.yml workflow files via make recompile to sync compiled output wi
#39010	Closed	Addresses the backlog of functions exceeding the 60-body-line lint limit in `pkg/workflow/
#39009	Closed	- [ ] Write Go AST rewriter tool to transform map[string]bool set patterns - [ ] Apply tra
#39008	Merged	upload_artifact was failing every safe_outputs job with HTTP 400 `malformed: the json
#39007	Merged	This change addresses nine context/environment handling findings where cancellation and ti
#38965	Merged	runs-on-slim was schema-validated and parsed as a string only, so workflows using the sa
#38953	Merged	gh-aw already writes GITHUB_AW_OTEL_TRACE_ID and GITHUB_AW_OTEL_PARENT_SPAN_ID to `GIT
#38949	Closed	The parser accepts both dispatch_repository and dispatch-repository spellings (documen
#38948	Closed	MCP metrics were modeled twice with incompatible duration types (string in audit models
#38947	Merged	gh aw env get and gh aw env update only exposed one-line summaries, leaving default fi
#38946	Merged	None of the ~36 cobra command definitions in pkg/cli/ used the Example: field — all co
#38945	Merged	pkg/workflow/mcp_github_config.go had ~19 functions accepting githubTool any that each
#38923	Merged	Compiler header now emits the auto-generated disclaimer as line 1 (before the ASCII logo)
#38918	Merged	When a workflow declares a top-level environment:, it was propagated to agent, `conclu
#38916	Merged	OTLP telemetry currently labels engine version as gh-aw.cli.version (and in `scope.versi
#38913	Merged	AWF command generation in gh-aw detected (redacted) Docker hosts for --docker-host-path-p
#38911	Open	ARC/DinD users currently need a bootstrap action to copy the copilot binary into the DinD

Notable PRs

Closed without merge ⚠️ (5)

• perf: fix 60% YAMLGeneration regression — skip yaml.Unmarshal for clean compiled workflows #39097 — perf: fix 60% YAMLGeneration regression (superseded)
• refactor(workflow): break down 10 largefunc violations in pkg/workflow #39010 — refactor: break down 10 largefunc violations in pkg/workflow
• [WIP] Refactor to replace map[string]bool with map[string]struct{} #39009 — [WIP] Refactor map[string]bool → map[string]struct{} (abandoned)
• fix(schema): add dispatch-repository dash alias to JSON schema #38949 — fix(schema): add dispatch-repository dash alias
• Unify MCP audit duration metrics to numeric milliseconds #38948 — Unify MCP audit duration metrics to numeric ms

Open PRs ⏳: 5 (#39130, #39118, #39100, #39089, #38911) — all opened within last 24h; none stale.

Key Insights

• Strong throughput, healthy merge rate: 39/49 merged (80%); completed-PR success is 89% (39/44). Work centered on the AI-credit/forecast guardrail subsystem and the safe-outputs MCP node runtime.
• Closures are mostly self-corrections, not rejections: 3 of 5 closed PRs were refactor/perf attempts (largefunc, struct{}, YAMLGeneration) the agent abandoned or superseded — worth watching if it persists.

---

Generated by Copilot Agent Analysis (Run: 27475626967)

Generated by 📊 Copilot Agent PR Analysis · 133.4 AIC · ⌖ 17.8 AIC · ⊞ 12.9K · ◷

• expires on Jun 14, 2026, 10:49 AM UTC-08:00