[prompt-clustering] Copilot agent prompt clusters — 2026-05-21 (n=1052, last 30 days)

[github-actions[bot]]

NLP clustering of copilot-swe-agent PR prompts from the last 30 days. TF-IDF features + K-means surfaced 9 task themes; one cluster ("awf / firewall / config") merges 20 pts below the 80.5% overall average and is the strongest candidate for prompt-engineering investment.

Summary

• Analysis Period: 2026-05-02 → 2026-05-21
• Total Tasks Analyzed: 1052
• Merged: 847 (80.5%)
• Closed (unmerged): 187
• Open: 18
• Clusters Identified: 9 (k chosen by silhouette)

Cluster Overview

#	Theme	Tasks	Share	Success	Avg Files	Avg +/-	Avg Commits
1	mcp / cli / added	362	34.4%	80.4%	22.7	+304.6/-173.5	4.11
2	workflow / agent / prompt	191	18.2%	80.6%	23.0	+370.9/-134.9	3.49
3	bug / fix / did	117	11.1%	82.9%	29.7	+515.5/-716.0	3.58
4	pr / comment / branch	103	9.8%	80.6%	16.1	+329.2/-74.4	5.28
5	otlp / engine / span	83	7.9%	81.9%	38.1	+355.0/-105.4	4.22
6	safe / output / outputs	76	7.2%	88.2%	29.7	+307.6/-137.7	5.55
7	awf / firewall / config	60	5.7%	60.0%	103.2	+930.8/-789.8	4.23
8	alias / model / inventory	33	3.1%	90.9%	33.3	+276.1/-67.8	3.24
9	actions / job / github actions	27	2.6%	77.8%	49.6	+254.7/-58.0	3.07

Cluster details, keywords, and examples

Cluster 1: mcp / cli / added

• Size: 362 tasks (34.4% of analyzed PRs)
• Success rate: 80.4% merged (291 merged, 59 closed, 12 open)
• Average change footprint: 22.7 files, +304.6/-173.5 lines, 4.11 commits, 2.53 comments
• Top keywords: mcp, cli, added, workflow, docs, updated, test, fix, behavior, shared
• Sample PRs:
  • ✅ #29860 — feat: migrate playwright workflows to CLI mode, deprecate MCP mode
  • ✅ #29888 — refactor: extract shared process_runner.cjs from claude and copilot harnesses
  • ✅ #29899 — fix: Pi engine uses COPILOT_GITHUB_TOKEN instead of PI_API_KEY
• Largest PRs in cluster:
  • ❌ #33128 — Drop shadow map[string]any fields: remove WorkflowData.Tools and ToolsConfig.raw
  • ❌ #32383 — Align compiler-generated step-name capitalization with glossary and Setup naming convention
  • ❌ #31136 — refactor: consolidate micro codemod files, update entity helpers, and generalize sortedMapKeys

Cluster 2: workflow / agent / prompt

• Size: 191 tasks (18.2% of analyzed PRs)
• Success rate: 80.6% merged (154 merged, 36 closed, 1 open)
• Average change footprint: 23.0 files, +370.9/-134.9 lines, 3.49 commits, 1.38 comments
• Top keywords: workflow, agent, prompt, experiment, token, daily, report, run, guidance, workflows
• Sample PRs:
  • ✅ #29843 — optimize agent-performance-analyzer with inline sub-agents
  • ✅ #29902 — fix(skill-optimizer): update workflow for v2.0.0 CLI interface
  • ✅ #29930 — feat(spec-librarian): add inline sub-agents for phases 1–3
• Largest PRs in cluster:
  • ✅ #32200 — Add emoji frontmatter field
  • ✅ #30995 — Import shared/observability-otlp.md in most agentic workflows
  • ✅ #33674 — Move Copilot /reflect fetch into agent prompt in daily-model-inventory workflow

Cluster 3: bug / fix / did

• Size: 117 tasks (11.1% of analyzed PRs)
• Success rate: 82.9% merged (97 merged, 20 closed, 0 open)
• Average change footprint: 29.7 files, +515.5/-716.0 lines, 3.58 commits, 1.75 comments
• Top keywords: bug, fix, bug fix, did, did fix, fix bug, workflow, testing, added, files
• Sample PRs:
  • ✅ #29936 — fix: harden isSafeExpression — refuse literal operands in &&/|| and fix spec-enforcer expression
  • ✅ #29949 — perf: fix FindIncludesInContent regression (~18% slower)
  • ✅ #29960 — perf: fix ExtractWorkflowNameFromFile regression — stream file instead of slurping
• Largest PRs in cluster:
  • ❌ #32254 — Recompile workflows: remove deprecated properties and regenerate lock files
  • ✅ #32193 — bump mcpg image to v0.3.9 (fixes wazero cache EACCES on artifact upload)
  • ❌ #32186 — fix(detection): download activation artifact before detection to resolve prompt.txt

Cluster 4: pr / comment / branch

• Size: 103 tasks (9.8% of analyzed PRs)
• Success rate: 80.6% merged (83 merged, 19 closed, 1 open)
• Average change footprint: 16.1 files, +329.2/-74.4 lines, 5.28 commits, 3.33 comments
• Top keywords: pr, comment, branch, run, review, workflow, reviewer, commit, merge, validation
• Sample PRs:
  • ✅ #29932 — fix(design-decision-gate): increase max-turns from 12 to 20
  • ✅ #30122 — Add mattpocock-skills-reviewer agentic workflow
  • ❌ #30211 — fix: increase design-decision-gate max-turns 12→15, add explicit push CLI syntax
• Largest PRs in cluster:
  • ✅ #33273 — Add on.pull_request_reviewer: slash_command synthetic trigger and reviewer lifecycle routing
  • ✅ #31614 — Auto-detect ARC/DinD and emit AWF --docker-host-path-prefix in generated workflows
  • ✅ #30909 — Fix github-script v8 fallback pin and annotate resolved/source fallback versions

Cluster 5: otlp / engine / span

• Size: 83 tasks (7.9% of analyzed PRs)
• Success rate: 81.9% merged (68 merged, 14 closed, 1 open)
• Average change footprint: 38.1 files, +355.0/-105.4 lines, 4.22 commits, 3.17 comments
• Top keywords: otlp, engine, span, spans, codemod, env, setup, values, shared, agent
• Sample PRs:
  • ✅ #29987 — feat: add gen_ai.usage token breakdown to conclusion spans
  • ✅ #30021 — feat: add support for multiple OTLP endpoints via polymorphic endpoint field
  • ✅ #30030 — fix: single-quote GH_AW_OTLP_ENDPOINTS to prevent YAML sequence parsing
• Largest PRs in cluster:
  • ✅ #32280 — Pass OTEL_EXPORTER_OTLP_HEADERS env var to mcpg container; remove headers from gateway JSON config
  • ✅ #31484 — Set frontmatter defaults and add shared import/expression support for max limits
  • ❌ #32298 — Propagate GH_AW_INFO_ENGINE_ID into setup steps so setup OTel spans emit gh-aw.engine.id

Cluster 6: safe / output / outputs

• Size: 76 tasks (7.2% of analyzed PRs)
• Success rate: 88.2% merged (67 merged, 8 closed, 1 open)
• Average change footprint: 29.7 files, +307.6/-137.7 lines, 5.55 commits, 3.66 comments
• Top keywords: safe, safe output, output, safe outputs, outputs, pr, chef, sous chef, sous, pr sous
• Sample PRs:
  • ✅ #29958 — fix: standardize safe outputs setup step names to use "Generate" verb consistently
  • ✅ #29999 — fix: smoke-copilot add_comment targets newly created discussion, not the one closed by close-older-discussions
  • ✅ #30071 — refactor: decouple safe-outputs checkout from event trigger context
• Largest PRs in cluster:
  • ✅ #32694 — Fix safe-outputs MCP startup crash from missing title-dedup modules
  • ✅ #32382 — Specify warn-mode threat-failure policies for all safe outputs
  • ✅ #32219 — Allow @copilot mentions in PR Sous Chef safe outputs

Cluster 7: awf / firewall / config

• Size: 60 tasks (5.7% of analyzed PRs)
• Success rate: 60.0% merged (36 merged, 22 closed, 2 open)
• Average change footprint: 103.2 files, +930.8/-789.8 lines, 4.23 commits, 6.18 comments
• Top keywords: awf, firewall, config, version, default, proxy, schema, gateway, workflow, golden
• Sample PRs:
  • ✅ #29848 — fix: version-pin AWF config $schema URL and add _schema field to JSONL types
  • ❌ #29859 — fix(gemini): route MCP gateway traffic through host.docker.internal when AWF firewall is enabled
  • ✅ #29962 — feat: add provider-prefix support to Pi engine (copilot/claude/codex routing)
• Largest PRs in cluster:
  • ✅ #31796 — Bump AWF to v0.25.44 and add firewall.effective-token-steering compiler support
  • ✅ #32161 — Remove firewall.effective-token-steering frontmatter key
  • ✅ #33200 — Expose reusable workflow network allowlist as an opt-in workflow_call input

Cluster 8: alias / model / inventory

• Size: 33 tasks (3.1% of analyzed PRs)
• Success rate: 90.9% merged (30 merged, 3 closed, 0 open)
• Average change footprint: 33.3 files, +276.1/-67.8 lines, 3.24 commits, 2.73 comments
• Top keywords: alias, model, inventory, docs, copilot, new, daily, added, entries, missing
• Sample PRs:
  • ✅ #29858 — Add model aliases and fallbacks to AWF config
  • ✅ #29984 — Add model alias format specification
  • ✅ #29995 — feat: add model alias validation enforcing the Model Alias Format spec
• Largest PRs in cluster:
  • ✅ #33661 — Update model alias inventory and ET multiplier registry for 2026-05-21
  • ✅ #33182 — Add per-engine default model aliases with Sonnet-sized agent defaults
  • ✅ #30739 — feat: model alias inventory update 2026-05-07

Cluster 9: actions / job / github actions

• Size: 27 tasks (2.6% of analyzed PRs)
• Success rate: 77.8% merged (21 merged, 6 closed, 0 open)
• Average change footprint: 49.6 files, +254.7/-58.0 lines, 3.07 commits, 0.26 comments
• Top keywords: actions, job, github actions, failing, fix, github, logs, url, description, root cause
• Sample PRs:
  • ✅ #30197 — fix: add actions: read permission to smoke-water.yml (#investigate-smoke-water-failure)
  • ✅ #32003 — [WIP] Fix failing GitHub Actions job lint-go
  • ✅ #32004 — [WIP] Fix failing GitHub Actions job Lint Gate
• Largest PRs in cluster:
  • ❌ #33673 — [WIP] Fix failing GitHub Actions job js-typecheck
  • ❌ #33672 — [WIP] Fix failing GitHub Actions job Lint Gate
  • ❌ #32190 — [WIP] Fix failing GitHub Actions job 'test'

Key Findings

1. Dominant task type: "mcp / cli / added" accounts for 34.4% of PRs (362 tasks) with a 80.4% merge rate.
2. Highest success rate: "alias / model / inventory" at 90.9% (30/33).
3. Lowest success rate: "awf / firewall / config" at 60.0% — this cluster is the strongest candidate for prompt-engineering improvements.
4. Largest change footprint: "awf / firewall / config" averages 103.2 files per PR — these tasks tend to span a lot of the codebase.
5. Most iterative: "safe / output / outputs" averages 5.55 commits per PR, suggesting more back-and-forth before the agent reached a final state.

Recommendations

1. Invest prompt tuning in "awf / firewall / config" — its 60.0% merge rate is well below the 80.5% overall average; clarifying acceptance criteria for this cluster has the highest leverage.
2. Templatize "mcp / cli / added" — it is the most common pattern (34.4% of work) and already merges at 80.4%; a reusable task template would lock in that quality.
3. Scope-check "awf / firewall / config" — averaging 103.2 files per PR is a risk for review burden; consider splitting these into smaller assignments before handing them to the agent.
4. Watch iteration in "safe / output / outputs" — 5.55 commits per PR suggests the agent is converging slowly; tightening the initial spec or adding examples could reduce drift.

Full PR cluster table

PR	Cluster	State	Files	+Adds	-Dels	Commits	Title
#33739	6	🕒 open	236	433	359	3	Harden aw-portfolio-yield finalization and add Sentry OTel fallback
#33716	6	🕒 open	6	45	76	4	Deduplicate workflow expression regex usage across extraction and validation pat
#33707	9	✅ merged	1	1	1	1	fix: add required_field_removals to toolsMeta JSDoc type
#33702	3	🕒 open	11	203	28	4	Fix pull_request_reviewer double-trigger on comment events
#33699	4	✅ merged	5	55	7	2	fix(model-inventory): 2026-05-21 — critical gpt-5.1-codex-mini multiplier fix +
#33694	6	✅ merged	9	60	32	2	perf: move regexp patterns to package-level variables (16 violations)
#33687	6	🕒 open	236	1080	536	6	refactor(create_pull_request): extract helpers module and add tests
#33685	6	🕒 open	20	75	74	6	Standardize pkg debug logging by removing log identifier collisions and stdlib
#33683	6	🕒 open	7	1375	1725	7	refactor(parser): break up oversized functions in pkg/parser to satisfy 60-line
#33676	6	✅ merged	4	109	37	8	Correct ET/ΔET token accounting for cached-input overlap in token_usage.jsonl
#33674	8	✅ merged	235	369	362	4	Move Copilot /reflect fetch into agent prompt in daily-model-inventory workflo
#33673	5	❌ closed	235	316	294	3	[WIP] Fix failing GitHub Actions job js-typecheck
#33672	5	❌ closed	234	294	294	3	[WIP] Fix failing GitHub Actions job Lint Gate
#33664	7	🕒 open	247	3379	3359	8	chore: bump default gh-aw-mcpg to v0.3.16 and gh-aw-firewall to v0.25.50
#33661	4	✅ merged	240	685	358	5	Update model alias inventory and ET multiplier registry for 2026-05-21
#33660	6	✅ merged	13	533	47	8	feat: support packaging action workflow files (.yml) in aw.yml
#33659	3	❌ closed	2	205	130	5	fix: push branch before creating fallback issue for protected-files fallback-to-
#33658	6	❌ closed	11	3424	674	5	Add engine.permission-mode as a first-class Claude engine config field
#33657	3	✅ merged	2	7	0	2	feat: PR triage agent reads customer triage rules from .github/triage.md at runt
#33656	6	✅ merged	2	113	11	6	fix(logs): resolve workflow display name when --repo is set and local lock files
#33655	8	✅ merged	2	75	127	3	contribution-check: offload report formatting and comment routing to small-model
#33654	2	✅ merged	4	27	3	3	fix: resolve workflow_install_note.md ENOENT in safe_outputs job
#33650	6	✅ merged	13	349	41	5	Support max-patch-size / max-patch-files under `safe-outputs.create-pull-req
#33647	6	✅ merged	2	5	5	2	Soften empty MCP RPC telemetry handling in gateway log parsing
#33646	6	✅ merged	87	192	105	2	Sync lock files with MinDiscussionBodyLength schema change; confirm formatting g
#33645	6	✅ merged	6	104	42	5	feat: complete x-deprecation-message coverage for all deprecated schema fields
#33639	2	✅ merged	18	453	40	8	Add body-allowed: false to close-discussion and close-issue safe-outputs
#33629	6	✅ merged	9	585	15	2	feat: show effective-token delta per MCP tool call in agent log
#33628	8	✅ merged	2	82	28	2	feat(token-usage): per-turn rows with ΔET and compounded ET in step summary
#33627	6	✅ merged	1	9	37	2	Enforce short-description punctuation rules across the full CLI command tree
#33626	6	✅ merged	1	150	127	6	Strengthen MCP add integration tests with testify assertions, table-driven cover
#33625	8	✅ merged	7	153	67	2	fix: set per-workflow token budgets and narrow file-glob patterns in meta-orches
#33623	1	✅ merged	2	1710	0	2	Add OTLP data quality validator workflow for end-to-end telemetry integrity chec
#33610	2	✅ merged	16	416	68	11	Support allowed-branches enforcement for create-pull-request safe output
#33596	8	✅ merged	7	34	5	2	Normalize report formatting guidelines across 7 agentic workflows
#33595	8	✅ merged	5	110	5	4	fix: guard create_discussion against PLACEHOLDER-only bodies
#33594	6	❌ closed	3	73	0	2	Align main workflow schema with parser allowlist by adding top-level applyTo a
#33591	6	✅ merged	5	130	2	5	Close daily SPDD spec gaps across MCP Scripts, Effective Tokens, Forecast, Front
#33575	6	❌ closed	0	0	0	3	Confirm CLI short-description punctuation consistency (no code delta)
#33574	6	❌ closed	2	99	7	10	Serialize create_issue handler execution to eliminate shared-state race condit
#33573	6	✅ merged	3	5	2	4	Harden Agentic Maintenance compile-workflows by removing Docker-only image valid
#33570	6	✅ merged	2	175	113	2	Update Daily OTel Advisor to use shared Sentry/Grafana OTEL MCP imports
#33562	3	✅ merged	30	118	75	11	fix: dispatch all pull_request_reviewer workflows from /review
#33553	2	✅ merged	67	262	261	3	Allow Chaos PR Bundle Fuzzer to modify tests/chaos/** in safe-output PR creati
#33552	3	❌ closed	66	244	244	3	Harden run step env map allocation against integer overflow in sanitizer
#33551	2	❌ closed	66	250	245	3	Guard safe-output event type merges against allocation size overflow
#33550	6	❌ closed	66	244	244	3	Harden generated workflow script against unsafe JSON quoting (CodeQL alert #600)
#33548	5	❌ closed	66	247	247	2	[WIP] Fix failing GitHub Actions job lint-go
#33543	6	✅ merged	3	4	2	2	Allow patch-diff.githubusercontent.com in the GitHub domain ecosystem
#33542	3	✅ merged	6	113	284	3	Remove centralized pull_request_reviewer dispatching from agentic_commands.yml
#33541	6	✅ merged	9	101	48	4	Improve linter analyzer ergonomics: ctxbackground autofix, test-file parity, and
#33540	8	✅ merged	2	196	97	2	Add sub_agent_strategy A/B experiment to smoke-gemini workflow
#33528	1	✅ merged	2	52	12	2	fix(otlp): always emit gen_ai.response.finish_reasons; use GITHUB_SHA as service
#33523	8	✅ merged	1	12	0	4	Improve Daily Reliability Review readability with progressive disclosure
#33510	1	🕒 open	2	176	6	2	fix(otlp): service.version fallback, timeout vs failure, finish_reasons, total_t
#33499	6	✅ merged	2	37	0	3	Strengthen shared Sentry prompts for grounded reliability triage
#33491	6	🕒 open	9	249	22	2	[plan] Migrate high-impact validation paths to NewValidationError with actiona
#33490	6	🕒 open	241	3948	3948	2	chore: bump default agentic CLI/MCP versions and regenerate compiled workflow ar
#33489	8	🕒 open	2	77	14	2	Add retry-blocked pre-activation guard to semantic function refactor workflow
#33482	6	🕒 open	6	27	42	3	refactor: extract renderDefaultJSONMCPConfig to eliminate duplicated MCP wrapper
#33467	6	🕒 open	11	769	103	4	feat: fuzzy "Did you mean?" suggestions for engine, event, permission, and MCP t
#33466	6	🕒 open	8	311	6	4	feat: add file/line context to validation errors
#33465	6	🕒 open	7	48	10	4	fix: handle errors from os.Setenv and os.Chdir in pkg/cli (10 sites)
#33464	6	✅ merged	6	111	7	7	feat: render reasoning/thinking messages with ◯ icon and italic styling
#33461	3	✅ merged	11	112	79	7	Remove pull_request_review from on.pull_request_reviewer hybrid routing
#33456	6	🕒 open	17	2625	2307	6	Remove per-server MCP network configuration
#33455	5	✅ merged	3	3	3	3	[WIP] Fix failing GitHub Actions job JS Tests (shard 2/4)
#33454	7	❌ closed	5	55	56	2	Migrate apiProxy basePath URL routes from CLI flags to JSON config
#33452	6	✅ merged	22	679	137	8	fix: resolve 12 lint violations — excess params, context.Background, os.Exit
#33451	6	❌ closed	14	215	7	6	fix: allow Copilot bot in mattpocock-skills-reviewer; handle GitHub App "not a u
#33450	6	❌ closed	9	57	56	2	Replace ambiguous log.* identifiers with explicit pkg/logger variables in li
#33449	3	✅ merged	2	24	1	4	Handle pull_request_review activation reactions in add_reaction
#33448	6	✅ merged	4	320	344	7	Refactor schedule parser long functions into focused helpers
#33447	3	✅ merged	3	34	1	3	Bypass slash-command gating for pull_request reviewer lifecycle events
#33441	6	✅ merged	2	82	6	3	Fix non-deterministic secret replacement in workflows with shared fallback expre
#33440	7	❌ closed	3	0	3	2	Codex: remove OPENAI_API_KEY requirement for AWF openai-proxy provider
#33435	3	✅ merged	3	12	7	2	Soft-skip submit_pull_request_review when no PR review context is available
#33434	6	✅ merged	21	108	198	7	refactor: semantic function clustering — dedup linter helpers, drop stub files,
#33433	6	✅ merged	6	66	18	4	[safeoutputs] Strengthen anti-probing guidance for add_comment and create_issue
#33432	6	✅ merged	4	29	18	2	Preserve JSON-imported workflow names in add/add-wizard resolution
#33431	6	✅ merged	3	143	174	4	[docs] Update Astro dependencies and apply Starlight 0.39 sidebar migration
#33430	4	✅ merged	5	12	2	3	Add missing gemini-3.5-flash ET multiplier to model inventory
#33429	5	✅ merged	3	2	3	3	[WIP] Fix failing GitHub Actions job lint-go
#33424	6	✅ merged	6	54	14	3	Polish JSON-import UX for add/add-wizard output
#33421	6	✅ merged	2	103	1	3	Prefer JSON title/name for URL imports when derived workflow name is GUID-like
#33420	6	✅ merged	2	33	18	3	Drop non-config JSON metadata during add/add-wizard import
#33419	6	✅ merged	2	3	5	2	fix: pass X-GitHub-Api-Version as top-level header in dispatchWorkflow
#33418	3	✅ merged	50	132	128	7	Skip unsupported pull_request_review events in activation reaction/status cond
#33412	3	✅ merged	8	265	75	3	Migrate ready-for-review agentic workflows to pull_request_reviewer
#33411	3	❌ closed	5	112	1	4	Document pull_request_reviewer trigger and centralized reviewer routing
#33402	6	✅ merged	2	89	27	5	Allow add URL importer auth for api.githubcopilot.com
#33386	6	❌ closed	7	16	117	15	Align stdio MCP network schema deprecation semantics with runtime proxy-args
#33385	1	✅ merged	1	31	0	3	Strengthen timeout_minutes codemod tests with multiple-occurrence coverage
#33384	6	✅ merged	16	162	126	3	Harden shared MCP allowlists for high-risk servers and document wildcard policy
#33368	6	✅ merged	3	17	7	5	Collapse generated footer install instructions behind details/summary disclosure
#33365	6	✅ merged	9	360	15	6	SPDD: close safeguards/consistency gaps across 5 specs and add targeted complian
#33363	8	✅ merged	2	87	255	4	Reduce CLI Consistency Checker token usage via pre-agent help capture and prompt
#33362	6	✅ merged	7	504	7	6	Add generic x-deprecation-message schema walker; emit warning when deprecated fr
#33361	3	✅ merged	2	166	0	7	Preserve inline review analysis in 422 body-only fallback
#33360	6	✅ merged	2	26	51	7	fix: remove orphan toolset (singular) deprecated property from schema
#33359	6	✅ merged	1	1	1	3	No-op: panic-recovery convention already restored in mcp_inspect_inspector gor
#33358	6	✅ merged	2	63	1	5	Treat zero-byte rpc-messages.jsonl as an MCP telemetry failure
#33353	6	✅ merged	6	224	2	7	Enforce compile-time rejection of queue: max with cancel-in-progress: true i
#33352	1	❌ closed	7	186	31	3	Add gh aw fix codemod for supportsLLMGateway → llmGatewayPort migration
#33350	2	🕒 open	18	524	56	16	Normalize safe-output issue/PR constraints to required-title-prefix, add migra
#33337	6	✅ merged	14	681	40	10	Prioritize compilation errors and add recovery guidance to gh aw compile
#33336	6	✅ merged	14	567	7	7	Add actionable error-message standards and changed-file custom linter
#33335	8	✅ merged	3	39	0	4	Normalize report-formatting guidance across reporting workflows
#33332	6	✅ merged	7	7	47	4	Harden MCP gh-aw binary discovery in shared setup import
#33328	6	✅ merged	5	27	23	3	Rename gitutil.ReadFileFromHEADWithRoot to ReadFileFromHEAD and align call s
#33327	6	✅ merged	3	14	1	3	Harden Agentic Maintenance compile-workflows with pre-compile validation
#33326	6	❌ closed	2	33	8	2	fix: CLI consistency improvements — lint --dir, logs --cache-before, forecast do
#33325	6	✅ merged	4	68	1	3	docs: add pkg/syncutil README and register dependency in cli/workflow specs
#33321	6	✅ merged	40	266	52	4	Fix MCP gh-aw binary discovery in Q dev install path
#33314	8	✅ merged	3	17	7	5	Harden daily experiment reporting with run→branch state verification
#33296	8	✅ merged	2	226	16	6	Add prompt_compression A/B experiment and caveman prompt variant to agent-pe
#33295	3	✅ merged	5	296	90	14	Merge main into PR branch, resolve workflow lock conflict, and regenerate comp
#33294	1	❌ closed	6	42	15	3	Add OTEL workflow/run attribute aliases for reliable token-consumption attributi
#33282	3	❌ closed	4	205	3	6	Enforce pre-push validation and retry-blocked branch checks for PR creation
#33281	6	✅ merged	2	3	3	4	[deps] Bump charmbracelet golden pseudo-version to 920740d613be
#33276	6	✅ merged	5	435	1	9	Adopt Lipgloss tree rendering for MCP hierarchy and status dependencies
#33275	4	✅ merged	4	87	15	6	Emit resolved runtime model as gen_ai.response.model on agent spans
#33273	3	✅ merged	257	1455	138	39	Add on.pull_request_reviewer: slash_command synthetic trigger and reviewer lif
#33272	6	✅ merged	29	292	292	4	Replace raw log.Print* usage in library packages with package-scoped logger in
#33271	8	❌ closed	67	611	635	2	[deep-report] Extract shared github-proxy-default import to DRY gh-proxy defau
#33250	4	✅ merged	8	163	15	6	Accept run_id alias for audit MCP tool parameters
#33249	6	✅ merged	7	145	39	4	Self-heal cache-memory git setup when restored metadata is corrupt
#33248	6	✅ merged	22	3229	88	2	Tighten golden-test coverage: remove dead console fixtures and goldenize wasm al
#33247	8	✅ merged	5	74	44	3	Reduce Step Name Alignment agent turns via deterministic pre-agent manifest
#33246	6	❌ closed	236	3934	3884	3	Bump default CLI/MCP toolchain versions and regenerate workflow locks
#33245	1	✅ merged	6	52	23	3	OTel: propagate GenAI semantic attrs to gh-aw.agent.conclusion spans
#33242	6	✅ merged	6	311	205	8	Refactor create-* config parsing to shared helper and unify string→array coercio
#33241	1	✅ merged	1	113	208	5	Refactor parser engine-include tests to table-driven testify assertions
#33240	6	✅ merged	44	506	458	16	Make setup-gh-aw install idempotent when gh-aw is already available
#33231	2	✅ merged	10	177	142	4	refactor: extract resolveSafeOutputIssueTarget to eliminate safe-output target r
#33221	8	✅ merged	2	66	26	3	Multi-Device Docs Tester: move Astro server startup to pre-agent steps to unbloc
#33220	8	✅ merged	6	45	38	2	Normalize report-style guidance across non-compliant issue/report workflows
#33219	7	🕒 open	8	103	17	2	Bind Node toolcache into AWF chroot for Copilot-engine workflow startup reliabil
#33218	8	✅ merged	1	10	0	2	[aw] Prevent Step Name Alignment from using invalid `gh search issues --state al
#33217	6	✅ merged	1	10	0	2	Add panic recovery to mcp inspect background goroutines
#33216	6	✅ merged	17	200	67	6	Enforce prod errstringmatch findings and centralize 403/410 error classification
#33205	5	✅ merged	6	26	26	3	[WIP] Fix failing GitHub Actions job 'test'
#33204	5	✅ merged	2	3	2	3	[WIP] Fix failing GitHub Actions job lint-go
#33203	5	✅ merged	6	26	26	3	[WIP] Fix failing GitHub Actions job build-wasm
#33202	1	✅ merged	6	60	30	4	Split shared OTLP backend import into Sentry, Grafana, and Datadog components
#33201	6	❌ closed	4	24	3	4	Replace ctx-local context.Background() fallbacks in action resolution and Dock
#33200	7	✅ merged	258	1562	771	13	Expose reusable workflow network allowlist as an opt-in workflow_call input
#33193	5	✅ merged	5	4	9	3	[WIP] Fix failing GitHub Actions job Lint Gate
#33192	2	❌ closed	11	569	597	5	Add orthogonal parameter aliases to safe output tool schemas
#33191	6	✅ merged	3	116	5	7	Validate required workflows on workflow_run during compile
#33182	4	✅ merged	239	769	716	6	Add per-engine default model aliases with Sonnet-sized agent defaults
#33181	6	✅ merged	7	216	11	12	Resolve #aw_* temporary IDs during bundle-based signed commit replay
#33180	6	✅ merged	5	162	47	4	Refactor current-repository lookup caches to shared syncutil.OnceLoader
#33179	8	✅ merged	3	71	19	3	Optimize CLI Consistency Checker via inline small-model sub-agents
#33178	3	✅ merged	4	174	3	6	Restore authenticated ls-remote in signed commit push path
#33177	4	✅ merged	7	24	3	2	Model inventory 2026-05-19: add raptor-mini alias coverage and missing GPT-5 s
#33171	2	✅ merged	5	75	2	4	Handle resolve_pull_request_review_thread integration-token 403 as actionable
#33170	2	✅ merged	6	173	15	7	fix(safe-outputs): warn agents about field-name divergence in add_labels, remove
#33168	6	✅ merged	36	36	36	2	Fix gh-aw install guard to handle github/gh-aw owner-prefix in `gh extension l
#33165	5	✅ merged	3	11	44	6	[WIP] Fix failing GitHub Actions job JS Tests (shard 4/4)
#33164	6	✅ merged	11	1137	13	9	feat: import workflow or JSON workflow from arbitrary HTTP(S) URL
#33163	3	✅ merged	3	71	2	4	fix: rename branch when recreate-ref deleteRef is blocked by branch protection r
#33152	8	✅ merged	2	19	1	3	Wrap experiment assignment summary in collapsible details block
#33144	5	✅ merged	3	87	39	3	[WIP] Fix failing GitHub Actions job 'Lint Gate'
#33134	2	✅ merged	5	206	5	11	Analyze: update_pull_request with empty args silently drops safe outputs
#33130	6	✅ merged	7	42	5	2	Align CLI help/docs consistency for outcomes, project setup, URLs, and wording
#33129	1	✅ merged	6	753	55	14	Fix compound || expressions in prompt markdown never substituting at runtime
#33128	6	❌ closed	303	2155	4097	5	Drop shadow map[string]any fields: remove WorkflowData.Tools and `ToolsConfi
#33127	3	❌ closed	2	41	1	2	fix: add scope-guard, acceptance criteria, and version-bump constraints to preve
#33125	3	✅ merged	2	215	86	2	feat(pr-sous-chef): run formatters and push to branch
#33124	7	✅ merged	233	2687	2682	4	Bump default AWF firewall to v0.25.49 and refresh compiled lockfiles
#33116	6	✅ merged	160	666	315	5	fix: use COPILOT_DUMMY_BYOK indirection to suppress secret-scanner false positiv
#33096	6	✅ merged	19	1157	3190	3	docs(spdd): strengthen safeguards, OTel norms, and compliance tests across 5 spe
#33093	6	✅ merged	7	290	81	4	Stop futile auth retries in Copilot/Claude/Codex harnesses after first-attempt a
#33089	2	✅ merged	2	602	15	10	Match agent failure issues by stored metadata instead of title alone
#33088	7	❌ closed	2	210	6	3	Add AWF installer fallback when pinned firewall release is missing
#33087	3	✅ merged	4	209	83	4	Render report_incomplete reasons in agent failure comment and issue templates
#33085	8	✅ merged	3	189	103	4	Trim token spend in Matt Pocock skills reviewer workflow
#33059	8	✅ merged	2	1593	0	3	feat: agentic workflow — rewrite merged PR descriptions in caveman style
#33050	3	✅ merged	2	53	0	1	Add pr-to-go-linter skill for PR-driven custom linter generation
#33044	2	✅ merged	14	635	123	14	Prevent safe-output PR probing from creating stray downstream pull requests
#33041	6	✅ merged	8	97	16	6	Support tools.github.allowed-repos: current for repo-scoped MCP guard policies
#33038	6	✅ merged	5	70	37	7	Refactor pkg mutex sites to use deferred unlocks consistently
#33037	1	✅ merged	2	114	1	2	fix(otlp): derive gh-aw.run.status and status.code from output errors when concl
#33036	8	✅ merged	2	166	111	2	Fix broken OTLP shared import in UK AI operational resilience workflow
#33035	8	✅ merged	2	23	17	2	fix(network): allow pypi.org and files.pythonhosted.org in daily-semgrep-scan
#33034	8	✅ merged	1	6	0	3	Add FAQ guidance for using skills with agentic workflows
#33033	8	✅ merged	16	395	9	15	Add github-app.missing-key ignore mode and guard App token minting across work
#33030	1	✅ merged	5	84	6	2	Add AWF/AWMG runtime versions to OTLP resource attributes
#33029	6	✅ merged	5	106	8	4	Prevent compile panic when workflow directory contains only shared components
#32944	7	✅ merged	236	3282	3287	3	Revert default firewall/MCP gateway bump from ac0fd25
#32939	4	✅ merged	2	52	31	2	fix(model-inventory): enrich /reflect null models via models_url fallback
#32938	8	❌ closed	3	30	80	3	fix(daily-model-inventory): remove runner-host /reflect pre-step and query refle
#32937	7	❌ closed	2	27	24	2	Pin Daily Compiler Threat Spec Optimizer to AWF v0.25.46
#32913	7	✅ merged	236	3287	3282	3	Bump default AWF firewall to v0.25.48 and MCP gateway to v0.3.11
#32911	6	✅ merged	12	1151	2	7	Add SSL Skill Normalizer and convert reporting, error-messages, jqschema to SSL
#32910	3	✅ merged	8	695	7	6	fix(safe-output): prevent silent 422 on PR review submission
#32909	6	✅ merged	3	33	88	4	Reject removed tools.serena in parser and align with schema
#32906	3	✅ merged	54	517	66	10	Harden privileged checkout path in q.lock.yml for comment-triggered runs
#32905	3	❌ closed	14	1914	186	10	Add BMAD-guided PR reviewer workflow and slash-command routing

Showing 200 most recent of 1052 analyzed PRs. Full CSV: pr-clusters.csv.

Methodology

• Extracted task description from PR title + body intro (text before ## Changes / ## Summary / similar) across 1054 PRs in cache.
• Cleaned markdown, code blocks, URLs; dropped PRs with <30 chars of cleaned text.
• TF-IDF features: vocab=300, unigrams + bigrams, min_df=5, max_df=0.6.
• K-means with k chosen by silhouette score on a 500-sample subset; k=9 won with score 0.029.
• Workflow turn/duration metrics were not available in this run (no aw_info.json logs cached); analysis uses PR-level proxies (commits, comments, file count) as iteration signals.
  References:
• Workflow run: §26221432692
• Cluster data CSV (in artifacts): pr-clusters.csv
• Cluster analysis JSON (in artifacts): cluster-analysis.json

Generated by 📊 Copilot Agent Prompt Clustering Analysis · ● 11.4M · ◷

• expires on May 22, 2026, 11:17 AM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Copilot Agent Prompt Clustering Analysis.

A newer discussion is available at Discussion #33980.