Analyze Sessions: Cross-Session Usage Report
Full-history review of local Claude and Codex prompt usage, token telemetry, recurring workflows, skill performance signals, and automation opportunities.
alignment_status: review generated: 2026-06-05 artifact: alignment/analyze-sessions-cross-session-usage.html
Overview Stats
Section feedback
This review page analyzes the default local Claude and Codex histories through June 5, 2026. It uses compact prompt histories for user-message counts, rich Codex rollout metadata for project and token enrichment, and repo lessons as corroborating evidence for recurring correction patterns.
16,609Total compact user prompts
6,377Compact sessions
2,685Codex rich rollout files
9,975,727,071Codex logged total tokens
| Source | Messages | Sessions | Date range | Active days | Share |
|---|---|---|---|---|---|
| claude | 10,053 | 3,811 | 2025-12-10 to 2026-06-05 | 127 | 60.5% |
| codex | 6,556 | 2,566 | 2026-01-20 to 2026-06-05 | 62 | 39.5% |
| Project | Messages | Share |
|---|---|---|
| bismarck-v0.4 | 1,438 | 8.66% |
| poke/monorepo | 1,284 | 7.73% |
| agentic-skills | 1,181 | 7.11% |
| lexcorp-war-room | 1,111 | 6.69% |
| metternich-engine | 1,084 | 6.53% |
| loadoutworks.com | 1,036 | 6.24% |
| bismarck-v0.3 | 929 | 5.59% |
| mobile-ideas | 537 | 3.23% |
| pitwall-monorepo | 513 | 3.09% |
| lexcorp-sdk | 416 | 2.5% |
| lexcorp-landing | 353 | 2.13% |
| pitwall | 348 | 2.1% |
| b4 | 345 | 2.08% |
| bismarck-v5 | 342 | 2.06% |
| claude-usage-review | 338 | 2.04% |
Source Comparison
Section feedback
Claude has the larger compact prompt archive and a longer date range. Codex has fewer compact prompts but richer per-session metadata and token telemetry. Recent Codex activity is concentrated in agentic-skills, while historical Claude activity is more distributed across application repos.
| Source | Top projects |
|---|---|
| claude |
|
| codex |
|
Recent 14-Day Trend
| Date | Claude prompts | Codex prompts |
|---|---|---|
| 2026-05-16 | 68 | 56 |
| 2026-05-17 | 110 | 110 |
| 2026-05-18 | 184 | 223 |
| 2026-05-19 | 87 | 39 |
| 2026-05-20 | 68 | 6 |
| 2026-05-21 | 110 | 127 |
| 2026-05-22 | 11 | 86 |
| 2026-05-23 | 5 | 10 |
| 2026-05-27 | 114 | 121 |
| 2026-05-28 | 82 | 56 |
| 2026-06-01 | 54 | 77 |
| 2026-06-03 | 51 | 61 |
| 2026-06-04 | 72 | 10 |
| 2026-06-05 | 32 | 3 |
Top Commands Overall
| Command | Count | Example |
|---|---|---|
| $run | 1,636 | $run |
| /ship | 1,541 | /ship --no-plan |
| /clear | 887 | /clear |
| /ship-end | 268 | /ship-end |
| /run | 182 | /run |
| /ship-then-plan | 161 | /ship-then-plan |
| $ship | 145 | $ship |
| /investigate | 99 | /investigate upon hitting enter after inputing my email within the upcoming events input form, it should not automatically subscribe without letting the user select from the drop... |
| /sync | 98 | /sync |
| $targeted-skill-builder | 94 | $targeted-skill-builder design-system benchmark timeout reporting |
| /usage | 89 | /usage |
| $benchmark-test-skill | 82 | $benchmark-test-skill design-system |
| $exec | 73 | [$exec](/Users/georgele/projects/tools/agentic-skills/packs/exec-loop/codex/exec/SKILL.md) |
| $ship-end | 70 | [$ship-end](/Users/georgele/projects/tools/claude-skills/codex/ship-end/SKILL.md) |
| /resume | 69 | /resume |
| $roadmap | 65 | [$roadmap](/Users/georgele/projects/tools/claude-skills/codex/roadmap/SKILL.md) |
| $session-triage | 62 | $session-triage design-system benchmark failure |
| $guide | 59 | $guide for getting all of the stripe information for all products in the lexcorp portfolio as specified by their documentation |
| $pack | 58 | $pack refresh |
| /roadmap | 54 | /roadmap |
Top Claude Commands
| Command | Count | Example |
|---|---|---|
| /ship | 1,540 | /ship --no-plan |
| /clear | 886 | /clear |
| /ship-end | 268 | /ship-end |
| /run | 182 | /run |
| /ship-then-plan | 161 | /ship-then-plan |
| /investigate | 99 | /investigate upon hitting enter after inputing my email within the upcoming events input form, it should not automatically subscribe without letting the user select from the drop... |
| /sync | 98 | /sync |
| /usage | 89 | /usage |
| /resume | 69 | /resume |
| /roadmap | 54 | /roadmap |
| /pack | 49 | /pack install business, game packs |
| /analyze-sessions | 35 | /analyze-sessions |
Top Codex Commands
| Command | Count | Example |
|---|---|---|
| $run | 1,635 | $run |
| $ship | 144 | $ship |
| $targeted-skill-builder | 94 | $targeted-skill-builder design-system benchmark timeout reporting |
| $benchmark-test-skill | 82 | $benchmark-test-skill design-system |
| $exec | 73 | [$exec](/Users/georgele/projects/tools/agentic-skills/packs/exec-loop/codex/exec/SKILL.md) |
| $ship-end | 70 | [$ship-end](/Users/georgele/projects/tools/claude-skills/codex/ship-end/SKILL.md) |
| $roadmap | 65 | [$roadmap](/Users/georgele/projects/tools/claude-skills/codex/roadmap/SKILL.md) |
| $session-triage | 62 | $session-triage design-system benchmark failure |
| $guide | 56 | $guide for getting all of the stripe information for all products in the lexcorp portfolio as specified by their documentation |
| $pack | 56 | $pack refresh |
| $sync | 41 | $sync |
| $benchmark-agent-review | 32 | $benchmark-agent-review benchmark-test-skill |
Token And Cost Check
Section feedback
Token totals are available only from Codex rich rollout token_count events. The aggregation uses the final or highest cumulative total_token_usage snapshot per session, avoiding cumulative event double-counting. Cached-input and reasoning-output fields are presented as logged subcategories, not added on top of total_tokens.
| Coverage field | Value |
|---|---|
| Codex rich sessions with metadata | 2,685 |
| Codex sessions with populated token totals | 2,633 |
| Token-count events | 125,560 |
| Populated token-count events | 123,188 |
| Claude usage metadata | unavailable in ~/.claude/history.jsonl |
| Codex compact usage metadata | unavailable in ~/.codex/history.jsonl; enriched from rich rollout token_count events where present |
| Token class | Logged total | Notes |
|---|---|---|
| Input tokens | 9,934,443,395 | Logged cumulative input tokens per session; includes cached input accounting in the provider telemetry shape. |
| Cached input tokens | 9,452,839,168 | Logged cached-input subcategory; not added separately to total tokens. |
| Output tokens | 41,283,676 | Logged output tokens. |
| Reasoning output tokens | 9,132,393 | Logged reasoning-output subcategory; not added separately to total tokens. |
| Total tokens | 9,975,727,071 | Final/highest cumulative total_token_usage snapshot per session, not a sum of every event. |
Top Token Projects
| Project | Total | Input | Cached input | Output | Reasoning output |
|---|---|---|---|---|---|
| agentic-skills | 1,643,526,627 | 1,638,622,360 | 1,575,583,872 | 4,904,267 | 898,025 |
| mobile-ideas | 1,498,370,660 | 1,492,493,517 | 1,434,706,048 | 5,877,143 | 697,905 |
| poke/monorepo | 698,547,483 | 695,136,763 | 662,876,416 | 3,410,720 | 1,095,561 |
| lexcorp-war-room | 530,318,796 | 528,268,389 | 496,294,912 | 2,050,407 | 571,559 |
| pitwall-monorepo | 469,816,877 | 467,989,853 | 440,838,400 | 1,827,024 | 310,145 |
| content | 433,715,515 | 432,156,191 | 412,290,944 | 1,559,324 | 235,310 |
| bismarck-v0.4 | 328,778,351 | 327,630,894 | 310,726,272 | 1,147,457 | 309,508 |
| lexcorp-sdk | 320,328,130 | 318,726,452 | 299,256,960 | 1,601,678 | 413,456 |
| claude-usage-review | 310,408,293 | 308,947,919 | 294,600,320 | 1,460,374 | 397,385 |
| metternich-engine | 307,372,445 | 306,077,250 | 289,040,384 | 1,295,195 | 515,804 |
Top Token Dates
| Date | Total | Input | Cached input | Output |
|---|---|---|---|---|
| 2026-04-16 | 725,115,166 | 721,924,167 | 693,512,064 | 3,190,999 |
| 2026-04-28 | 685,798,757 | 682,599,130 | 648,357,248 | 3,199,627 |
| 2026-05-18 | 578,915,217 | 577,533,294 | 559,454,848 | 1,381,923 |
| 2026-05-15 | 529,923,182 | 528,180,596 | 505,390,336 | 1,742,586 |
| 2026-05-13 | 471,616,314 | 470,079,972 | 451,095,936 | 1,536,342 |
| 2026-05-14 | 427,805,947 | 426,354,526 | 408,203,520 | 1,451,421 |
| 2026-05-12 | 393,076,819 | 391,877,265 | 377,591,808 | 1,199,554 |
| 2026-04-21 | 358,331,351 | 356,838,257 | 336,947,200 | 1,493,094 |
| 2026-04-19 | 346,932,104 | 345,310,443 | 329,609,600 | 1,621,661 |
| 2026-04-22 | 321,294,563 | 320,222,431 | 305,819,648 | 1,072,132 |
Top Token Sessions
| Date | Project | Total | Input | Output | Provider | Model |
|---|---|---|---|---|---|---|
| 2026-05-12 | content | 145,208,784 | 144,762,388 | 446,396 | openai | not_logged |
| 2026-05-10 | jit-life-os | 98,619,816 | 98,356,919 | 262,897 | openai | not_logged |
| 2026-04-22 | mobile-ideas | 80,212,773 | 80,118,263 | 94,510 | openai | not_logged |
| 2026-04-30 | mobile-ideas | 65,468,846 | 65,207,578 | 261,268 | openai | not_logged |
| 2026-05-10 | content | 44,204,896 | 44,031,799 | 173,097 | openai | not_logged |
| 2026-04-21 | bismarck-v0.4 | 42,176,207 | 42,082,453 | 93,754 | openai | not_logged |
| 2026-04-09 | metternich-engine | 41,567,098 | 41,398,559 | 168,539 | openai | not_logged |
| 2026-05-18 | bismarck-v5 | 36,600,086 | 36,552,217 | 47,869 | openai | not_logged |
| 2026-04-21 | mobile-ideas | 34,552,204 | 34,478,559 | 73,645 | openai | not_logged |
| 2026-05-05 | mobile-ideas | 31,390,746 | 31,253,233 | 137,513 | openai | not_logged |
Cost status: cost unavailable: no direct structured cost fields found, no user-provided pricing table, and rich session metadata does not log model identifiers needed for a reliable current pricing map. Pricing source: none used. The report does not estimate dollars from remembered pricing.
Categorized Patterns
Section feedback
The largest recurring categories are general short instructions, shipping/commit boundaries, skill-pack maintenance, and step execution. General small instructions are kept separate so they do not drown out actionable automation opportunities.
| Category | Messages | Source split |
|---|---|---|
| General small instruction / answer | 7,040 | claude: 4427, codex: 2613 |
| Ship existing work / commit-push boundary | 3,380 | claude: 2911, codex: 469 |
| Execute next planned step | 2,589 | claude: 660, codex: 1929 |
| Skill and pack maintenance | 1,276 | claude: 278, codex: 998 |
| Docs, specs, and task planning | 655 | claude: 465, codex: 190 |
| Debugging, tests, and verification | 622 | claude: 483, codex: 139 |
| Research, investigation, and reports | 518 | claude: 402, codex: 116 |
| Sync and environment setup | 331 | claude: 283, codex: 48 |
| Frontend, UX, prototype, and animation | 198 | claude: 144, codex: 54 |
Fuzzy Prompt Patterns
| Pattern | Count | Examples |
|---|---|---|
| Ship existing work / commit-push boundary | 3,380 |
|
| Execute next planned step | 2,781 |
|
| Skill and pack maintenance | 3,284 |
|
| Debugging, tests, and verification | 1,067 |
|
| Research, investigation, and reports | 1,123 |
|
| Frontend, UX, prototype, and animation | 480 |
|
| Docs, specs, and task planning | 1,889 |
|
| Sync and environment setup | 652 |
|
Exact Repeated Prompts
| Count | Sources | Example |
|---|---|---|
| 1,627 | claude: 1, codex: 1626 | $run |
| 1,534 | claude: 1534 | /ship |
| 887 | claude: 886, codex: 1 | /clear |
| 264 | claude: 264 | /ship-end |
| 216 | claude: 169, codex: 47 | continue |
| 177 | claude: 145, codex: 32 | yes please |
| 175 | claude: 175 | /run |
| 161 | claude: 161 | /ship-then-plan |
| 146 | claude: 142, codex: 4 | commit and push |
| 143 | claude: 1, codex: 142 | $ship |
| 98 | claude: 98 | /sync |
| 89 | claude: 89 | /usage |
| 76 | claude: 75, codex: 1 | update docs, commit and push |
| 72 | codex: 72 | [$exec](/Users/georgele/projects/tools/agentic-skills/packs/exec-loop/codex/exec/SKILL.md) |
| 67 | codex: 67 | Implement the plan. |
Workflow Sequences
Section feedback
Common sequences show clear workflow migration: Claude sessions often use slash-command shipping and clearing loops, while Codex sessions show benchmark/triage/remediation chains and dollar-command task execution.
| Sequence | Count | Example source |
|---|---|---|
| /ship-end -> /clear | 123 | claude session 628cf529-8e81-401c-960f-2f7b91951d5b |
| /ship -> /clear | 111 | claude session 4b9bc8f7-303f-4302-a5dd-e2eaf359e672 |
| $session-triage -> $targeted-skill-builder | 52 | codex session 019e1302-6655-78f3-a084-c9d2dc618f15 |
| $benchmark-test-skill -> $session-triage | 43 | codex session 019e12f6-145e-78b3-910a-01fbe1616510 |
| $benchmark-test-skill -> $session-triage -> $targeted-skill-builder | 38 | codex session 019e1302-6655-78f3-a084-c9d2dc618f15 |
| $run -> $ship | 35 | codex session 019d4e94-ba0b-70d3-9f21-a0a52a4d8d8e |
| /run -> /ship | 27 | claude session 08d1a86b-27de-4071-97bf-04e88a817ddb |
| /roadmap -> /clear | 24 | claude session 4fa784c4-a013-4d00-b835-021d5eb800d6 |
| $benchmark-test-skill -> $benchmark-agent-review | 24 | codex session 019e1d57-bdd7-7832-b884-69d44b812da0 |
| /sync -> /ship | 22 | claude session 4fa784c4-a013-4d00-b835-021d5eb800d6 |
| $benchmark-agent-review -> $targeted-skill-builder | 22 | codex session 019e1d57-bdd7-7832-b884-69d44b812da0 |
| /sync -> /clear | 21 | claude session 4696c70d-decf-455e-8f8a-addf063f501b |
| /ship-end -> /ship | 20 | claude session 11a09fae-036f-47a8-8980-2a00dd5a0f70 |
| $benchmark-test-skill -> $benchmark-agent-review -> $targeted-skill-builder | 18 | codex session 019e1d57-bdd7-7832-b884-69d44b812da0 |
| /ship -> /ship-end | 17 | claude session 63a114a1-aa71-449a-a3ff-6147812f080b |
Skill Performance Trends
Section feedback
Skill performance evidence combines compact prompt usage, repo prompt-history files, and lessons. The lessons are especially important because they encode user corrections and the concrete prevention rule that should be automated.
Prompt History Counts
| Prompt history slug | Files |
|---|---|
| ship | 45 |
| analyze-sessions | 28 |
| investigate | 24 |
| ship-end | 20 |
| exec | 20 |
| pack | 17 |
| session-triage | 4 |
| reconcile-dev-docs | 4 |
| sync | 4 |
| targeted-skill-builder | 4 |
| devtool-docs-audit | 3 |
| benchmark-test-skill | 3 |
| feature-interview | 2 |
| research-roadmap | 2 |
| provision-agentic-config | 2 |
Recent Lessons
| Date | Lesson title |
|---|---|
| 2026-06-02 | Do not turn deploy-skill absence into next work when deploy is not contextually needed |
| 2026-06-01 | Verify stale unchecked tasks before recommending next work |
| 2026-05-31 | Alignment page compile controls must stay in normal page flow |
| 2026-05-31 | Alignment feedback YAML needs a local section path |
| 2026-05-30 | Never write a test/verification claim from remembered or garbled tool output |
| 2026-05-30 | Prove a test failure is pre-existing before calling it unrelated |
| 2026-05-30 | Verify commit/push RC and the staged boundary, don't assume |
| 2026-05-30 | Alignment feedback needs an early YAML path |
| 2026-05-29 | Status routing must follow canonical skill contracts |
| 2026-05-28 | Codex `$` visibility can fail on stale symlink skill installs |
Correction Signals From Prompt History
Correction/frustration regex matches: Claude 152; Codex 134. These are approximate signals, not a judgment of fault; examples below are sanitized excerpts.
| Source | Date | Project | Example |
|---|---|---|---|
| claude | 2025-12-10 | walkthrough-starting-point | can you try npm run build again? |
| claude | 2025-12-10 | loadoutworks.com | woah wait why did you make that change to the index? everything in the old code works |
| claude | 2025-12-12 | loadoutworks.com | can we provide support for right-side view as well? I don't think top view would be often used. I might be wrong but that's my personal feeling |
| claude | 2025-12-12 | loadoutworks.com | no the left image is 1950 x 570 and the right image is 1940x494. can you check the image dimensions again and verify your recommendation? |
| claude | 2025-12-19 | loadoutworks.com | try again? |
| claude | 2026-01-08 | bismarck-v0.3 | how do I check to see if the orchestrator is working? I get this error now: Failed to load sessions: Service temporarily unavailable. Please try again later. |
| claude | 2026-01-09 | bismarck-v0.3 | can you check the runners and orchestrator? bc things have timed out again |
| claude | 2026-01-09 | bismarck-v0.3 | Ok the terminal is popping back up again, but we still have no interactivity. I put the logs in logs.md. You might to check on the runners and orchestrator |
| claude | 2026-01-09 | bismarck-v0.3 | please try again with the previous commands, I configured the sso for you |
| claude | 2026-01-10 | bismarck-v0.3 | can you note that down in the claude.md so we don't make that mistake again? |
| claude | 2026-01-13 | bismarck-v0.3 | finish step 1 and stop |
| claude | 2026-01-15 | bismarck-v0.3 | ok, looks like everything works on production. Redirect is still broken due to the redirect uri and I have confirmed that the callback url has been updated correctly |
Evidence Matrix And Assumptions
Section feedback
| Claim | Evidence | Inference | Confidence | Assumption status | Decision impact |
|---|---|---|---|---|---|
| Histories were readable and parsed completely | Source inventory: Claude 10,053 lines; Codex compact 6,556 lines; Codex rich 2,685 files / 854,835 records; parse_errors = 0. | The analysis can rely on complete local source coverage for the default scope. | High | Assumes current file contents at analysis time are the desired scope. | Approve or correct the scope before treating the report as complete. |
| Codex has usable token metadata; Claude compact history does not | Codex rich logs: 123,188 populated token_count events and 2,633 token sessions. Claude history has prompt records only. | Token totals are reportable for Codex rich sessions only; source-level token comparison against Claude is not possible from these logs. | High | Token telemetry shape is trusted as logged. | Use token findings for Codex spend patterns; do not compare to Claude spend. |
| Cost is unavailable | No direct structured cost fields were found; model identifiers are not logged in session_meta; no user-provided pricing table was supplied. | Estimating dollars would require guessing model mapping or fetching current pricing, so the report stops at token totals. | High | A future run with verified pricing and model IDs could estimate cost. | Approve cost as unavailable or request a pricing-table-backed rerun. |
| Recent Codex work is dominated by skills/pack maintenance | Activity category count: Skill and pack maintenance = 1,276; top Codex project = agentic-skills with 715 messages. | Codex is being used heavily for repository workflow automation and skill maintenance rather than only app feature work. | High | Category classification is regex-based and may overcount prompts mentioning skill terms incidentally. | Prioritize automation around skill maintenance and shipping boundaries. |
| Shipping and verification loops are high risk and frequent | Top commands include /ship 1,541, $ship 145, /ship-end 268; lessons include verification-claim and stale-task failures. | The strongest durable workflow gap is not raw implementation, but proving and shipping work without stale claims or stale todo routing. | High | Lesson file is repo evidence, not a prompt-history source; used as corroboration for skill performance trends. | Review the proposed targeted-skill-builder route. |
| Plan-step execution remains a repeat workflow | Execute next planned step pattern count = 2,589; top command $run = 1,636. | A phase executor/runner remains a high-frequency workflow that benefits from guardrails and task-document synchronization. | Medium | Counts include explicit command invocations and natural-language step requests. | Consider integrating task execution and verification gates. |
Confidence and Assumption Register
| Item | Status | What would change it |
|---|---|---|
| Prompt counts | Evidence-backed from compact histories | New history lines written after this run will change counts. |
| Codex project attribution | Evidence-backed when session_id mapped to rich session_meta | Unknown Codex compact sessions or missing metadata would reduce attribution coverage. |
| Token totals | Evidence-backed for Codex rich sessions with populated cumulative totals | Provider telemetry changes or model-specific billing details would change interpretation, not logged token counts. |
| Cost | Unavailable, not estimated | A user-supplied or currently verified pricing table plus model IDs would enable estimation. |
| Automation frequencies | Provisional classification based on regex and command extraction | A semantic classifier or manual taxonomy could adjust category counts. |
| Recommended route | Inference from frequencies plus repo lessons | User priority may favor skill-maintenance batcher over shipping verifier despite the verification risk. |
Coverage Gaps
- Claude history does not contain token usage fields, so Claude token and cost comparisons are unavailable.
- Codex compact history does not contain project/cost/token data by itself; project and token attribution depends on matching rich session metadata.
- Codex session metadata logs provider as
openaibut does not log model names; cost estimation is intentionally blocked without reliable model mapping. - Workflow categories are regex-classified and should be treated as directional evidence; command counts are exact after normalizing literal commands and Codex markdown skill links.
Ranked Recommendations
Section feedback
The highest-leverage automations are the ones that remove repeated multi-step prompt chains while reducing known failure modes from the lessons file.
| Rank | Pattern | Frequency | Type | Suggested automation |
|---|---|---|---|---|
| 1 | Skill/pack maintenance dominates recent Codex work | 3,284 | Skill | Batch skill version/archive/changelog/parity/generator validation into a single repo-managed maintenance skill |
| 2 | Validation/ship loops repeatedly ask for prompt capture, task updates, validation, commit, and push | 3,021 | Skill contract hardening | Add an always-on shipping boundary checklist with machine-readable verification claim capture |
| 3 | Repeated plan-step execution prompts across projects | 2,781 | Agent or skill | Standardize phase executor that reads tasks/todo, executes the next unchecked step, verifies, and records review notes |
| 4 | Correction lessons show routing/next-command drift and stale task docs | 74 | Standing instruction plus tests | Add next-route normalization and stale-task artifact reconciliation checks to shipping/status skills |
| 5 | Alignment pages recur across research/report workflows | 55 | Plugin/integration or template | Reusable alignment-page generator with schema-driven gates and validation checks |
Top 5 Highest-Impact Automations
| Automation | Avoided manual prompts estimate | Why it matters |
|---|---|---|
| Skill maintenance batcher | 3,284 | Skill versioning, archive, changelog, mirror parity, prompt history, and showcase validation are a repeated multi-step sequence. |
| Shipping boundary verifier | 3,021 | Commit/push/status requests are frequent and risky when verification claims are copied from scrollback. |
| Task phase executor | 2,781 | Many prompts ask to implement or move to a specific roadmap/todo step. |
| Correction-to-lesson gate | 286 | Corrections already produce durable lessons, but a gate could detect and record the lesson before final output. |
| Alignment page generator/validator | 55 | Alignment-page construction and feedback controls are cross-cutting and easy to drift. |
Proposed next route for approval:
Likely owner surface: ship/ship-end/codebase-status/analyze-sessions contracts plus layer1 route/verification tests. Validation expectation: focused layer1 tests that fail on unverified pass/fail claims and stale todo-only next-work recommendations.
$targeted-skill-builder enforce shipping verification claim capture and stale-task reconciliation gateLikely owner surface: ship/ship-end/codebase-status/analyze-sessions contracts plus layer1 route/verification tests. Validation expectation: focused layer1 tests that fail on unverified pass/fail claims and stale todo-only next-work recommendations.
Review Gates
Section feedback
Answer these gates only after reviewing the evidence and recommendations above. Feedback-only YAML can be compiled from any section feedback control before answering every required gate.
Research Completeness
Is the evidence sufficient to accept this cross-session analysis as the current report?
Assumptions/Confidence
How should token and cost limitations be handled?
Recommended Path
Which proposed automation direction should be treated as the next approved path?
Artifact Destination
Is the review artifact/provenance destination acceptable?
Post-Approval Route
After final approval, what should the agent do next?
Compile YAML
Section feedback
Use feedback YAML for emphasis requests, concerns, or clarification before answering all gates. Use final answers YAML only when every required gate is answered.