From c6ee987bb59c329e84e4c6ace205c9078049a16c Mon Sep 17 00:00:00 2001 From: yaojingang Date: Sun, 14 Jun 2026 09:29:05 +0800 Subject: [PATCH] feat: surface deferred resource context cost --- .../reports/context_budget.json | 34 +++++- .../reports/context_budget.json | 34 +++++- registry/index.json | 2 +- registry/packages/yao-meta-skill.json | 2 +- reports/context_budget.json | 34 +++++- reports/context_budget.md | 10 +- reports/context_budget_summary.json | 16 ++- reports/registry_audit.json | 4 +- reports/registry_audit.md | 2 +- reports/review-studio.html | 14 +-- reports/review-studio.json | 114 +++++++++++++++--- reports/security_trust_report.json | 2 +- reports/security_trust_report.md | 2 +- reports/skill-overview.json | 6 +- reports/skill_os2_audit.json | 2 +- reports/skill_os2_audit.md | 2 +- reports/skill_os2_coverage.json | 2 +- reports/skill_os2_coverage.md | 2 +- reports/upgrade_check.json | 2 +- scripts/render_context_reports.py | 12 +- scripts/render_review_studio.py | 8 +- scripts/resource_boundary_check.py | 38 ++++++ scripts/review_studio_gates.py | 11 +- tests/verify_quality_checks.py | 9 ++ tests/verify_review_studio.py | 38 +++++- 25 files changed, 337 insertions(+), 65 deletions(-) diff --git a/examples/complex-release-orchestrator/generated-skill/reports/context_budget.json b/examples/complex-release-orchestrator/generated-skill/reports/context_budget.json index 5c28725..d6e09de 100644 --- a/examples/complex-release-orchestrator/generated-skill/reports/context_budget.json +++ b/examples/complex-release-orchestrator/generated-skill/reports/context_budget.json @@ -6,9 +6,39 @@ "context_budget_tier": "production", "context_budget_limit": 1000, "skill_body_tokens": 718, - "other_text_tokens": 1946, + "other_text_tokens": 2113, "estimated_initial_load_tokens": 790, - "estimated_total_text_tokens": 2664, + "estimated_total_text_tokens": 2831, + "deferred_resource_tokens": 1657, + "deferred_resource_warn_threshold": 120000, + "deferred_resource_dirs": [ + { + "path": "references", + "estimated_tokens": 554, + "file_count": 4 + }, + { + "path": "scripts", + "estimated_tokens": 367, + "file_count": 1 + }, + { + "path": "input", + "estimated_tokens": 315, + "file_count": 1 + }, + { + "path": "outputs", + "estimated_tokens": 263, + "file_count": 1 + }, + { + "path": "evals", + "estimated_tokens": 158, + "file_count": 1 + } + ], + "large_deferred_resource_dirs": [], "relevant_file_count": 12, "unused_resource_dirs": [], "quality_signal_points": 130, diff --git a/examples/governed-incident-command/generated-skill/reports/context_budget.json b/examples/governed-incident-command/generated-skill/reports/context_budget.json index 98b114e..199f26c 100644 --- a/examples/governed-incident-command/generated-skill/reports/context_budget.json +++ b/examples/governed-incident-command/generated-skill/reports/context_budget.json @@ -6,9 +6,39 @@ "context_budget_tier": "production", "context_budget_limit": 1000, "skill_body_tokens": 658, - "other_text_tokens": 1603, + "other_text_tokens": 1770, "estimated_initial_load_tokens": 760, - "estimated_total_text_tokens": 2261, + "estimated_total_text_tokens": 2428, + "deferred_resource_tokens": 1030, + "deferred_resource_warn_threshold": 120000, + "deferred_resource_dirs": [ + { + "path": "scripts", + "estimated_tokens": 278, + "file_count": 1 + }, + { + "path": "references", + "estimated_tokens": 266, + "file_count": 3 + }, + { + "path": "outputs", + "estimated_tokens": 203, + "file_count": 1 + }, + { + "path": "input", + "estimated_tokens": 153, + "file_count": 1 + }, + { + "path": "evals", + "estimated_tokens": 130, + "file_count": 1 + } + ], + "large_deferred_resource_dirs": [], "relevant_file_count": 13, "unused_resource_dirs": [], "quality_signal_points": 130, diff --git a/registry/index.json b/registry/index.json index 58dc5ab..22796dc 100644 --- a/registry/index.json +++ b/registry/index.json @@ -16,7 +16,7 @@ "vscode" ], "package_metadata": "registry/packages/yao-meta-skill.json", - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" } ] } diff --git a/registry/packages/yao-meta-skill.json b/registry/packages/yao-meta-skill.json index 0ae2496..de4e867 100644 --- a/registry/packages/yao-meta-skill.json +++ b/registry/packages/yao-meta-skill.json @@ -16,7 +16,7 @@ "trust_level": "local", "license": "MIT", "checksums": { - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb", + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5", "archive_sha256": "76793bd367e9aaf6c25e6b11ac1139156950827bd87ba3dd8a1bdc883f7d70fe" }, "compatibility": { diff --git a/reports/context_budget.json b/reports/context_budget.json index 8ac8fd1..056e08a 100644 --- a/reports/context_budget.json +++ b/reports/context_budget.json @@ -1,14 +1,42 @@ { "ok": true, "failures": [], - "warnings": [], + "warnings": [ + "Deferred resource footprint is high: 381942 estimated tokens across references/scripts/evals. Keep Review Studio warnings visible until the largest resource dirs are split, archived, or justified." + ], "stats": { "context_budget_tier": "production", "context_budget_limit": 1000, "skill_body_tokens": 751, - "other_text_tokens": 1183524, + "other_text_tokens": 1187833, "estimated_initial_load_tokens": 944, - "estimated_total_text_tokens": 1184275, + "estimated_total_text_tokens": 1188584, + "deferred_resource_tokens": 381942, + "deferred_resource_warn_threshold": 120000, + "deferred_resource_dirs": [ + { + "path": "scripts", + "estimated_tokens": 334459, + "file_count": 95 + }, + { + "path": "references", + "estimated_tokens": 25648, + "file_count": 32 + }, + { + "path": "evals", + "estimated_tokens": 21835, + "file_count": 27 + } + ], + "large_deferred_resource_dirs": [ + { + "path": "scripts", + "estimated_tokens": 334459, + "file_count": 95 + } + ], "relevant_file_count": 500, "unused_resource_dirs": [], "quality_signal_points": 130, diff --git a/reports/context_budget.md b/reports/context_budget.md index a4f7dcc..f219ab0 100644 --- a/reports/context_budget.md +++ b/reports/context_budget.md @@ -1,9 +1,9 @@ # Context Budget Summary -| Target | Path | Tier | Limit | Initial | SKILL | Quality Density | Unused Dirs | Status | -| --- | --- | --- | ---: | ---: | ---: | ---: | --- | --- | -| root | `.` | `production` | 1000 | 944 | 751 | 137.7 | - | ok | -| complex-release-orchestrator | `examples/complex-release-orchestrator/generated-skill` | `production` | 1000 | 790 | 718 | 164.6 | - | ok | -| governed-incident-command | `examples/governed-incident-command/generated-skill` | `production` | 1000 | 760 | 658 | 171.1 | - | ok | +| Target | Path | Tier | Limit | Initial | SKILL | Deferred | Large Deferred Dirs | Quality Density | Unused Dirs | Status | +| --- | --- | --- | ---: | ---: | ---: | ---: | --- | ---: | --- | --- | +| root | `.` | `production` | 1000 | 944 | 751 | 381942 | scripts:334459 | 137.7 | - | ok | +| complex-release-orchestrator | `examples/complex-release-orchestrator/generated-skill` | `production` | 1000 | 790 | 718 | 1657 | - | 164.6 | - | ok | +| governed-incident-command | `examples/governed-incident-command/generated-skill` | `production` | 1000 | 760 | 658 | 1030 | - | 171.1 | - | ok | Per-target JSON reports are written beside each target and under `reports/`. diff --git a/reports/context_budget_summary.json b/reports/context_budget_summary.json index 18c1b52..b3110b3 100644 --- a/reports/context_budget_summary.json +++ b/reports/context_budget_summary.json @@ -8,10 +8,20 @@ "budget_limit": 1000, "initial_tokens": 944, "skill_body_tokens": 751, + "deferred_resource_tokens": 381942, + "large_deferred_resource_dirs": [ + { + "path": "scripts", + "estimated_tokens": 334459, + "file_count": 95 + } + ], "quality_density": 137.7, "unused_resource_dirs": [], "ok": true, - "warnings": [] + "warnings": [ + "Deferred resource footprint is high: 381942 estimated tokens across references/scripts/evals. Keep Review Studio warnings visible until the largest resource dirs are split, archived, or justified." + ] }, { "label": "complex-release-orchestrator", @@ -20,6 +30,8 @@ "budget_limit": 1000, "initial_tokens": 790, "skill_body_tokens": 718, + "deferred_resource_tokens": 1657, + "large_deferred_resource_dirs": [], "quality_density": 164.6, "unused_resource_dirs": [], "ok": true, @@ -32,6 +44,8 @@ "budget_limit": 1000, "initial_tokens": 760, "skill_body_tokens": 658, + "deferred_resource_tokens": 1030, + "large_deferred_resource_dirs": [], "quality_density": 171.1, "unused_resource_dirs": [], "ok": true, diff --git a/reports/registry_audit.json b/reports/registry_audit.json index 9197912..c94cfea 100644 --- a/reports/registry_audit.json +++ b/reports/registry_audit.json @@ -21,7 +21,7 @@ "trust_level": "local", "license": "MIT", "checksums": { - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb", + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5", "archive_sha256": "76793bd367e9aaf6c25e6b11ac1139156950827bd87ba3dd8a1bdc883f7d70fe" }, "compatibility": { @@ -78,7 +78,7 @@ "vscode" ], "package_metadata": "registry/packages/yao-meta-skill.json", - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" } ] }, diff --git a/reports/registry_audit.md b/reports/registry_audit.md index 777807a..4998b71 100644 --- a/reports/registry_audit.md +++ b/reports/registry_audit.md @@ -6,7 +6,7 @@ - Maturity: `governed` - Owner: `Yao Team` - License: `MIT` -- Package SHA256: `c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb` +- Package SHA256: `2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5` - Archive SHA256: `76793bd367e9aaf6c25e6b11ac1139156950827bd87ba3dd8a1bdc883f7d70fe` - Install simulated: `True` diff --git a/reports/review-studio.html b/reports/review-studio.html index 8d61ee4..e48daa5 100644 --- a/reports/review-studio.html +++ b/reports/review-studio.html @@ -428,7 +428,7 @@
审查结论 review - Score 91/100 + Score 89/100
@@ -439,17 +439,17 @@

审查闸门

-
通过

意图画布

intent confidence 100/100; Intent is clear enough to package the first routeable version.

reports/intent-confidence.json 证据
通过

触发实验

13 trigger cases; 0 misroutes; 0 ambiguous

reports/route_scorecard.json 证据
关注

输出实验

5/5 cases; with-skill 100.0; baseline 0.0; file-backed 1; near-neighbor 1; blind A/B 5; exec 10; command 10; model 0; recorded 0; reviewed 0/5; review pending 5

reports/output_quality_scorecard.json 证据
通过

上下文

initial load 944/1000; quality density 137.7

reports/context_budget.json 证据
通过

运行矩阵

5 / 5 targets pass

reports/conformance_matrix.json 证据
通过

信任报告

0 secrets; 95 scripts; 3 network-capable scripts; 0 help smoke failures

reports/security_trust_report.json 证据
通过

Python 兼容

Python 3.11; 157 files; 0 compatibility issues; 0 syntax; 0 f-string 3.11 hazards

reports/python_compatibility.json 证据
通过

架构维护

154 Python files; 0 hotspots; 0 blockers; largest 899 lines; 34 CLI handlers

reports/architecture_maintainability.json 证据
通过

权限批准

3/3 permissions approved; gaps 0; required file_write, network, subprocess

reports/security_trust_report.json + security/permission_policy.json 证据
通过

权限探针

4/4 targets probed; native 0; metadata fallback 4; installer 4; residual risks 4

reports/runtime_permission_probes.json 证据
通过

组合治理

12 skills, 1 actionable; 0 actionable route collisions; 0 actionable owner gaps; 0 actionable stale; 0 actionable drift; 24 scoped non-actionable issues

reports/skill_atlas.json 证据
通过

运营回路

1 metadata events; adoption 100.0; missed 0; bad-output 0; risk low

reports/adoption_drift_report.json 证据
关注

人工批准

0 active waivers; 1 warning gates still need reviewer decision

reports/review_waivers.json 证据
关注

世界证据

4 pending world-class evidence entries; 1 human pending; 3 external pending; overclaim guard true

reports/world_class_evidence_ledger.json 证据
通过

注册审计

yao-meta-skill 1.1.0; 6/6 compatibility entries pass; install pass with 4 adapters; installer permissions 12 enforced / 0 failures

reports/registry_audit.json + reports/install_simulation.json 证据
通过

发布路线

0 promote; 3 keep current; 0 blocked; upgrade minor declared / minor recommended

reports/promotion_decisions.json + reports/upgrade_check.json + docs/migration-v2.md 证据
+
通过

意图画布

intent confidence 100/100; Intent is clear enough to package the first routeable version.

reports/intent-confidence.json 证据
通过

触发实验

13 trigger cases; 0 misroutes; 0 ambiguous

reports/route_scorecard.json 证据
关注

输出实验

5/5 cases; with-skill 100.0; baseline 0.0; file-backed 1; near-neighbor 1; blind A/B 5; exec 10; command 10; model 0; recorded 0; reviewed 0/5; review pending 5

reports/output_quality_scorecard.json 证据
关注

上下文

initial load 944/1000; deferred 381942/120000; top deferred scripts 334459; quality density 137.7

reports/context_budget.json 证据
通过

运行矩阵

5 / 5 targets pass

reports/conformance_matrix.json 证据
通过

信任报告

0 secrets; 95 scripts; 3 network-capable scripts; 0 help smoke failures

reports/security_trust_report.json 证据
通过

Python 兼容

Python 3.11; 157 files; 0 compatibility issues; 0 syntax; 0 f-string 3.11 hazards

reports/python_compatibility.json 证据
通过

架构维护

154 Python files; 0 hotspots; 0 blockers; largest 899 lines; 34 CLI handlers

reports/architecture_maintainability.json 证据
通过

权限批准

3/3 permissions approved; gaps 0; required file_write, network, subprocess

reports/security_trust_report.json + security/permission_policy.json 证据
通过

权限探针

4/4 targets probed; native 0; metadata fallback 4; installer 4; residual risks 4

reports/runtime_permission_probes.json 证据
通过

组合治理

12 skills, 1 actionable; 0 actionable route collisions; 0 actionable owner gaps; 0 actionable stale; 0 actionable drift; 24 scoped non-actionable issues

reports/skill_atlas.json 证据
通过

运营回路

1 metadata events; adoption 100.0; missed 0; bad-output 0; risk low

reports/adoption_drift_report.json 证据
关注

人工批准

0 active waivers; 2 warning gates still need reviewer decision

reports/review_waivers.json 证据
关注

世界证据

4 pending world-class evidence entries; 1 human pending; 3 external pending; overclaim guard true

reports/world_class_evidence_ledger.json 证据
通过

注册审计

yao-meta-skill 1.1.0; 6/6 compatibility entries pass; install pass with 4 adapters; installer permissions 12 enforced / 0 failures

reports/registry_audit.json + reports/install_simulation.json 证据
通过

发布路线

0 promote; 3 keep current; 0 blocked; upgrade minor declared / minor recommended

reports/promotion_decisions.json + reports/upgrade_check.json + docs/migration-v2.md 证据

阻断事项

无。

-

关注事项

  • 输出实验5/5 cases; with-skill 100.0; baseline 0.0; file-backed 1; near-neighbor 1; blind A/B 5; exec 10; command 10; model 0; recorded 0; reviewed 0/5; review pending 5
  • 人工批准0 active waivers; 1 warning gates still need reviewer decision
  • 世界证据4 pending world-class evidence entries; 1 human pending; 3 external pending; overclaim guard true
+

关注事项

  • 输出实验5/5 cases; with-skill 100.0; baseline 0.0; file-backed 1; near-neighbor 1; blind A/B 5; exec 10; command 10; model 0; recorded 0; reviewed 0/5; review pending 5
  • 上下文initial load 944/1000; deferred 381942/120000; top deferred scripts 334459; quality density 137.7
  • 人工批准0 active waivers; 2 warning gates still need reviewer decision
  • 世界证据4 pending world-class evidence entries; 1 human pending; 3 external pending; overclaim guard true

修复动作

-
关注

输出实验

补足 output eval 覆盖、execution evidence、blind A/B 和 reviewer adjudication。

没有输出质量和人工盲评证据时，Skill 只能证明会触发，不能证明输出真的更好且经得起审查。
修复位置
evals/output/cases.jsonl + reports/output_quality_scorecard.md + reports/output_review_kit.html + reports/output_review_adjudication.md
验证命令
python3 scripts/adjudicate_output_review.py --write-template && python3 scripts/yao.py output-review
关注

人工批准

对保留的 warning 写入 reviewer、理由、范围和到期时间，或修掉 warning。

warning 可以被接受，但必须可审计、会过期，并且不能掩盖 blocker。
修复位置
reports/review_waivers.md
验证命令
python3 scripts/render_review_waivers.py .
关注

世界证据

补齐 provider、真人盲评、原生权限执行和真实客户端遥测证据，或明确本次发布不声明 world-class 完成。

世界级结论必须来自已接受的外部/人工证据；计划、metadata fallback、待评审和本地命令都不能替代完成证据。
修复位置
reports/world_class_operator_runbook.html + reports/world_class_evidence_ledger.md + reports/world_class_evidence_intake.md + reports/world_class_submission_review.md
验证命令
python3 scripts/yao.py world-class-runbook . && python3 scripts/yao.py world-class-ledger . && python3 scripts/yao.py review-studio .
+
关注

输出实验

补足 output eval 覆盖、execution evidence、blind A/B 和 reviewer adjudication。

没有输出质量和人工盲评证据时，Skill 只能证明会触发，不能证明输出真的更好且经得起审查。
修复位置
evals/output/cases.jsonl + reports/output_quality_scorecard.md + reports/output_review_kit.html + reports/output_review_adjudication.md
验证命令
python3 scripts/adjudicate_output_review.py --write-template && python3 scripts/yao.py output-review
关注

上下文

压缩或拆分高成本 deferred resources，保留最小可路由上下文。

初始加载可以安全，但后续 references、scripts、evals 体量过大时，reviewer 仍需要看到维护和读取成本。
修复位置
SKILL.md + references/ + scripts/ + evals/
验证命令
python3 scripts/render_context_reports.py
关注

人工批准

对保留的 warning 写入 reviewer、理由、范围和到期时间，或修掉 warning。

warning 可以被接受，但必须可审计、会过期，并且不能掩盖 blocker。
修复位置
reports/review_waivers.md
验证命令
python3 scripts/render_review_waivers.py .
关注

世界证据

补齐 provider、真人盲评、原生权限执行和真实客户端遥测证据，或明确本次发布不声明 world-class 完成。

世界级结论必须来自已接受的外部/人工证据；计划、metadata fallback、待评审和本地命令都不能替代完成证据。
修复位置
reports/world_class_operator_runbook.html + reports/world_class_evidence_ledger.md + reports/world_class_evidence_intake.md + reports/world_class_submission_review.md
验证命令
python3 scripts/yao.py world-class-runbook . && python3 scripts/yao.py world-class-ledger . && python3 scripts/yao.py review-studio .
@@ -501,12 +501,12 @@
-

上下文

initial load 944/1000; quality density 137.7

+

上下文

initial load 944/1000; deferred 381942/120000; top deferred scripts 334459; quality density 137.7

编译证据

Review reports/compiled_targets.md before packaging to inspect target adapter modes, generated files, preserved semantics, warnings, and unsupported features.

-

信任报告

Secret
0
脚本数
95
网络脚本
3
Help 失败
0
包体哈希
c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb
+

信任报告

Secret
0
脚本数
95
网络脚本
3
Help 失败
0
包体哈希
2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5

安全边界

高风险 secret、远程 inline execution、缺失依赖策略或无法解释的脚本接口应阻断 governed release。

@@ -536,7 +536,7 @@
-

人工批准

0 active waivers; 1 warning gates still need reviewer decision

+

人工批准

0 active waivers; 2 warning gates still need reviewer decision

批准台账

Waiver Count
0
Active Count
0
Expired Count
0
Invalid Count
0
覆盖 Gate
0
diff --git a/reports/review-studio.json b/reports/review-studio.json index 55c3c35..9b46e71 100644 --- a/reports/review-studio.json +++ b/reports/review-studio.json @@ -4,11 +4,11 @@ "skill_dir": ".", "summary": { "decision": "review", - "world_class_score": 91, + "world_class_score": 89, "gate_count": 16, "blocker_count": 0, - "warning_count": 3, - "action_count": 3, + "warning_count": 4, + "action_count": 4, "annotation_count": 0, "open_annotation_count": 0, "open_annotation_blocker_count": 0, @@ -42,8 +42,8 @@ { "key": "context-budget", "label": "上下文", - "status": "pass", - "detail": "initial load 944/1000; quality density 137.7", + "status": "warn", + "detail": "initial load 944/1000; deferred 381942/120000; top deferred scripts 334459; quality density 137.7", "evidence": "reports/context_budget.json", "link": "context_budget.md" }, @@ -115,7 +115,7 @@ "key": "review-waivers", "label": "人工批准", "status": "warn", - "detail": "0 active waivers; 1 warning gates still need reviewer decision", + "detail": "0 active waivers; 2 warning gates still need reviewer decision", "evidence": "reports/review_waivers.json", "link": "review_waivers.md" }, @@ -154,11 +154,19 @@ "evidence": "reports/output_quality_scorecard.json", "link": "output_quality_scorecard.md" }, + { + "key": "context-budget", + "label": "上下文", + "status": "warn", + "detail": "initial load 944/1000; deferred 381942/120000; top deferred scripts 334459; quality density 137.7", + "evidence": "reports/context_budget.json", + "link": "context_budget.md" + }, { "key": "review-waivers", "label": "人工批准", "status": "warn", - "detail": "0 active waivers; 1 warning gates still need reviewer decision", + "detail": "0 active waivers; 2 warning gates still need reviewer decision", "evidence": "reports/review_waivers.json", "link": "review_waivers.md" }, @@ -234,6 +242,52 @@ "evidence_link": "output_quality_scorecard.md", "verification_command": "python3 scripts/adjudicate_output_review.py --write-template && python3 scripts/yao.py output-review" }, + { + "gate_key": "context-budget", + "label": "上下文", + "status": "warn", + "priority": "warning", + "summary": "压缩或拆分高成本 deferred resources，保留最小可路由上下文。", + "why": "初始加载可以安全，但后续 references、scripts、evals 体量过大时，reviewer 仍需要看到维护和读取成本。", + "source_fix": "SKILL.md + references/ + scripts/ + evals/", + "source_refs": [ + { + "path": "SKILL.md", + "label": "entrypoint", + "kind": "source", + "line": 9, + "exists": true, + "link": "../SKILL.md" + }, + { + "path": "reports/context_budget.md", + "label": "context budget", + "kind": "report", + "line": 1, + "exists": true, + "link": "context_budget.md" + }, + { + "path": "scripts/resource_boundary_check.py", + "label": "resource boundary checker", + "kind": "source", + "line": 44, + "exists": true, + "link": "../scripts/resource_boundary_check.py" + }, + { + "path": "references/skill-engineering-method.md", + "label": "skill engineering method", + "kind": "method", + "line": 208, + "exists": true, + "link": "../references/skill-engineering-method.md" + } + ], + "evidence": "reports/context_budget.json", + "evidence_link": "context_budget.md", + "verification_command": "python3 scripts/render_context_reports.py" + }, { "gate_key": "review-waivers", "label": "人工批准", @@ -4996,7 +5050,7 @@ "label": "Review Studio", "status": "pass", "objective": "One HTML page supports first-pass production review across trigger, output, runtime, trust, release, and evidence actions.", - "current": "16 gates; decision review; warnings 3", + "current": "16 gates; decision review; warnings 4", "command": "python3 scripts/yao.py review-studio .", "test": "python3 tests/verify_review_studio.py", "evidence": [ @@ -10257,7 +10311,7 @@ "interactive_script_count": 0, "package_hash_scope": "source-contract-without-generated-reports", "package_hash_file_count": 179, - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" }, "failures": [], "warnings": [], @@ -13646,7 +13700,7 @@ "architecture_maintainability": { "schema_version": "1.0", "ok": true, - "generated_at": "2026-06-13", + "generated_at": "2026-06-14", "skill_dir": ".", "summary": { "python_file_count": 154, @@ -13758,14 +13812,42 @@ "context_budget": { "ok": true, "failures": [], - "warnings": [], + "warnings": [ + "Deferred resource footprint is high: 381942 estimated tokens across references/scripts/evals. Keep Review Studio warnings visible until the largest resource dirs are split, archived, or justified." + ], "stats": { "context_budget_tier": "production", "context_budget_limit": 1000, "skill_body_tokens": 751, - "other_text_tokens": 1183524, + "other_text_tokens": 1187832, "estimated_initial_load_tokens": 944, - "estimated_total_text_tokens": 1184275, + "estimated_total_text_tokens": 1188583, + "deferred_resource_tokens": 381942, + "deferred_resource_warn_threshold": 120000, + "deferred_resource_dirs": [ + { + "path": "scripts", + "estimated_tokens": 334459, + "file_count": 95 + }, + { + "path": "references", + "estimated_tokens": 25648, + "file_count": 32 + }, + { + "path": "evals", + "estimated_tokens": 21835, + "file_count": 27 + } + ], + "large_deferred_resource_dirs": [ + { + "path": "scripts", + "estimated_tokens": 334459, + "file_count": 95 + } + ], "relevant_file_count": 500, "unused_resource_dirs": [], "quality_signal_points": 130, @@ -17237,7 +17319,7 @@ "trust_level": "local", "license": "MIT", "checksums": { - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb", + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5", "archive_sha256": "76793bd367e9aaf6c25e6b11ac1139156950827bd87ba3dd8a1bdc883f7d70fe" }, "compatibility": { @@ -17294,7 +17376,7 @@ "vscode" ], "package_metadata": "registry/packages/yao-meta-skill.json", - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" } ] }, @@ -18330,7 +18412,7 @@ { "field": "package_sha256", "from": "0000000000000000000000000000000000000000000000000000000000000000", - "to": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "to": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" } ] }, diff --git a/reports/security_trust_report.json b/reports/security_trust_report.json index bcabbe8..1fe5194 100644 --- a/reports/security_trust_report.json +++ b/reports/security_trust_report.json @@ -23,7 +23,7 @@ "interactive_script_count": 0, "package_hash_scope": "source-contract-without-generated-reports", "package_hash_file_count": 179, - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" }, "failures": [], "warnings": [], diff --git a/reports/security_trust_report.md b/reports/security_trust_report.md index 47ec8a5..cfdecac 100644 --- a/reports/security_trust_report.md +++ b/reports/security_trust_report.md @@ -16,7 +16,7 @@ - Interactive scripts: `0` - Package hash scope: `source-contract-without-generated-reports` - Package hash files: `179` -- Package SHA256: `c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb` +- Package SHA256: `2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5` ## Failures diff --git a/reports/skill-overview.json b/reports/skill-overview.json index 72be843..996b7ac 100644 --- a/reports/skill-overview.json +++ b/reports/skill-overview.json @@ -1046,7 +1046,7 @@ "interactive_script_count": 0, "package_hash_scope": "source-contract-without-generated-reports", "package_hash_file_count": 179, - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" }, "skill_atlas": { "skill_count": 12, @@ -1084,7 +1084,7 @@ "trust_level": "local", "license": "MIT", "checksums": { - "package_sha256": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb", + "package_sha256": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5", "archive_sha256": "76793bd367e9aaf6c25e6b11ac1139156950827bd87ba3dd8a1bdc883f7d70fe" }, "compatibility": { @@ -1216,7 +1216,7 @@ { "field": "package_sha256", "from": "0000000000000000000000000000000000000000000000000000000000000000", - "to": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "to": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" } ] }, diff --git a/reports/skill_os2_audit.json b/reports/skill_os2_audit.json index a96090b..a561010 100644 --- a/reports/skill_os2_audit.json +++ b/reports/skill_os2_audit.json @@ -307,7 +307,7 @@ "key": "review-studio", "label": "Review Studio", "status": "pass", - "current": "decision review; warnings 3; score 91", + "current": "decision review; warnings 4; score 89", "target": "One page shows gates, evidence paths, blockers, warnings, actions, waivers, and annotations", "evidence": [ { diff --git a/reports/skill_os2_audit.md b/reports/skill_os2_audit.md index b385bb1..564bbdc 100644 --- a/reports/skill_os2_audit.md +++ b/reports/skill_os2_audit.md @@ -28,7 +28,7 @@ Generated at: `2026-06-13` | Native Permission Enforcement | external-required | native-enforced targets 0; installer-enforced targets 4 | At least one target/client enforces approved permissions at runtime | Integrate a real target-client or external installer runtime guard before claiming native permission enforcement. | | Skill Atlas | pass | 12 skills; actionable collisions 0 | Workspace catalog, route overlap, stale/owner gaps, drift, and no-route opportunities | Feed real drift data into Atlas once client telemetry is installed. | | Registry Distribution | pass | zip entries 574; install failures 0; permission failures 0 | Package metadata, archive checksum, package verification, and install simulation pass | Regenerate registry after package verification so checksums stay aligned. | -| Review Studio | pass | decision review; warnings 3; score 91 | One page shows gates, evidence paths, blockers, warnings, actions, waivers, and annotations | Resolve human/external warning gates before claiming full release readiness. | +| Review Studio | pass | decision review; warnings 4; score 89 | One page shows gates, evidence paths, blockers, warnings, actions, waivers, and annotations | Resolve human/external warning gates before claiming full release readiness. | | Telemetry Drift | pass | events 1; risk low; recipes 5 | Local-first metadata-only event contract, aggregate drift report, hook recipes, and import path | Keep raw JSONL out of distributed packages and use aggregate reports for Atlas. | | Native Client Telemetry | external-required | external source events 0; adoption samples 1 | A real Browser/Chrome/provider client sends production metadata events | Install a real client against the native host and import production metadata-only events. | diff --git a/reports/skill_os2_coverage.json b/reports/skill_os2_coverage.json index 45e9736..8f28c02 100644 --- a/reports/skill_os2_coverage.json +++ b/reports/skill_os2_coverage.json @@ -237,7 +237,7 @@ "label": "Review Studio", "status": "pass", "objective": "One HTML page supports first-pass production review across trigger, output, runtime, trust, release, and evidence actions.", - "current": "16 gates; decision review; warnings 3", + "current": "16 gates; decision review; warnings 4", "command": "python3 scripts/yao.py review-studio .", "test": "python3 tests/verify_review_studio.py", "evidence": [ diff --git a/reports/skill_os2_coverage.md b/reports/skill_os2_coverage.md index 2a7ef08..72c0379 100644 --- a/reports/skill_os2_coverage.md +++ b/reports/skill_os2_coverage.md @@ -24,7 +24,7 @@ This report maps the Skill OS 2.0 upgrade blueprint to concrete local artifacts, | Trust Security | `pass` | 95 scripts; secrets 0; help failures 0 | `python3 scripts/yao.py trust .` | `python3 tests/verify_trust_check.py` | | Skill Atlas | `pass` | 12 scanned skills; actionable collisions 0 | `python3 scripts/yao.py skill-atlas --workspace-root .` | `python3 tests/verify_skill_atlas.py` | | Registry Distribution | `pass` | archive entries 574; install failures 0 | `python3 scripts/yao.py package . --platform openai --platform claude --platform generic --platform vscode --output-dir dist --zip && python3 scripts/yao.py registry-audit .` | `python3 tests/verify_registry_audit.py` | -| Review Studio | `pass` | 16 gates; decision review; warnings 3 | `python3 scripts/yao.py review-studio .` | `python3 tests/verify_review_studio.py` | +| Review Studio | `pass` | 16 gates; decision review; warnings 4 | `python3 scripts/yao.py review-studio .` | `python3 tests/verify_review_studio.py` | | Telemetry Drift | `pass` | events 1; recipes 5; risk low | `python3 scripts/yao.py telemetry-hooks . && python3 scripts/yao.py adoption-drift .` | `python3 tests/verify_telemetry_hooks.py` | ## Recommended PR Coverage diff --git a/reports/upgrade_check.json b/reports/upgrade_check.json index 4b80c22..ef6354a 100644 --- a/reports/upgrade_check.json +++ b/reports/upgrade_check.json @@ -75,7 +75,7 @@ { "field": "package_sha256", "from": "0000000000000000000000000000000000000000000000000000000000000000", - "to": "c42643ea854dd90bb03ca607595f6a6b2c00118da64a78e8bdbb4392d7f204fb" + "to": "2d237fe645ef197282d514782f3bd37025c00e4d7efda725145aa5c3af180ba5" } ] }, diff --git a/scripts/render_context_reports.py b/scripts/render_context_reports.py index ea57191..4274745 100644 --- a/scripts/render_context_reports.py +++ b/scripts/render_context_reports.py @@ -52,6 +52,8 @@ def main() -> None: "budget_limit": stats.get("context_budget_limit"), "initial_tokens": stats.get("estimated_initial_load_tokens"), "skill_body_tokens": stats.get("skill_body_tokens"), + "deferred_resource_tokens": stats.get("deferred_resource_tokens"), + "large_deferred_resource_dirs": stats.get("large_deferred_resource_dirs", []), "quality_density": stats.get("quality_density"), "unused_resource_dirs": stats.get("unused_resource_dirs", []), "ok": report.get("ok"), @@ -72,14 +74,18 @@ def main() -> None: lines = [ "# Context Budget Summary", "", - "| Target | Path | Tier | Limit | Initial | SKILL | Quality Density | Unused Dirs | Status |", - "| --- | --- | --- | ---: | ---: | ---: | ---: | --- | --- |", + "| Target | Path | Tier | Limit | Initial | SKILL | Deferred | Large Deferred Dirs | Quality Density | Unused Dirs | Status |", + "| --- | --- | --- | ---: | ---: | ---: | ---: | --- | ---: | --- | --- |", ] for row in rows: unused = ", ".join(row["unused_resource_dirs"]) if row["unused_resource_dirs"] else "-" + large_dirs = ", ".join( + f"{item['path']}:{item['estimated_tokens']}" + for item in row.get("large_deferred_resource_dirs", []) + ) or "-" status = "ok" if row["ok"] else "fail" lines.append( - f"| {row['label']} | `{row['path']}` | `{row['budget_tier']}` | {row['budget_limit']} | {row['initial_tokens']} | {row['skill_body_tokens']} | {row['quality_density']} | {unused} | {status} |" + f"| {row['label']} | `{row['path']}` | `{row['budget_tier']}` | {row['budget_limit']} | {row['initial_tokens']} | {row['skill_body_tokens']} | {row['deferred_resource_tokens']} | {large_dirs} | {row['quality_density']} | {unused} | {status} |" ) lines.extend(["", "Per-target JSON reports are written beside each target and under `reports/`."]) (ROOT / "reports" / "context_budget.md").write_text("\n".join(lines) + "\n", encoding="utf-8") diff --git a/scripts/render_review_studio.py b/scripts/render_review_studio.py index e214e8b..5148d1b 100644 --- a/scripts/render_review_studio.py +++ b/scripts/render_review_studio.py @@ -154,12 +154,12 @@ ACTION_GUIDANCE: dict[str, dict[str, str]] = { "verification": "python3 scripts/adjudicate_output_review.py --write-template && python3 scripts/yao.py output-review", }, "context-budget": { - "summary": "压缩入口与高成本 references，保留最小可路由上下文。", - "why": "上下文成本过高会降低加载稳定性，并挤压用户任务材料的预算。", - "source_fix": "SKILL.md + references/", - "source_paths": [ + "summary": "压缩或拆分高成本 deferred resources，保留最小可路由上下文。", "why": "初始加载可以安全，但后续 references、scripts、evals 体量过大时，reviewer 仍需要看到维护和读取成本。", + "source_fix": "SKILL.md + references/ + scripts/ + evals/", "source_paths": [ {"path": "SKILL.md", "label": "entrypoint", "kind": "source", "patterns": ["# Yao Meta Skill"]}, {"path": "reports/context_budget.md", "label": "context budget", "kind": "report", "patterns": ["# Context"]}, + {"path": "scripts/resource_boundary_check.py", "label": "resource boundary checker", "kind": "source", "patterns": ["DEFERRED_RESOURCE"]}, + {"path": "references/skill-engineering-method.md", "label": "skill engineering method", "kind": "method", "patterns": ["Design Principle"]}, ], "verification": "python3 scripts/render_context_reports.py", }, diff --git a/scripts/resource_boundary_check.py b/scripts/resource_boundary_check.py index d0c5b41..b1065b6 100644 --- a/scripts/resource_boundary_check.py +++ b/scripts/resource_boundary_check.py @@ -41,6 +41,9 @@ CONTEXT_BUDGETS = { } SKILL_BODY_BUFFER = 100 SKILL_BODY_WARN_RATIO = 0.85 +DEFERRED_RESOURCE_DIRS = {"references", "scripts", "evals", "templates", "assets", "input", "outputs"} +DEFERRED_RESOURCE_WARN_TOKENS = 120_000 +DEFERRED_RESOURCE_WARN_DIR_TOKENS = 80_000 def has_files(path: Path) -> bool: @@ -128,6 +131,8 @@ def analyze_skill( other_tokens = 0 initial_load_tokens = 0 total_text_tokens = 0 + deferred_resource_tokens = 0 + deferred_resource_dirs: dict[str, dict[str, int | str]] = {} for path in files: if path.suffix and path.suffix not in TEXT_EXTS and path.name != "SKILL.md": continue @@ -135,6 +140,7 @@ def analyze_skill( tokens = estimate_tokens(text) total_text_tokens += tokens rel = path.relative_to(root) + top_dir = rel.parts[0] if rel.parts else str(rel) if rel == Path("SKILL.md"): skill_body_tokens += tokens initial_load_tokens += tokens @@ -142,6 +148,14 @@ def analyze_skill( other_tokens += tokens if rel.parts[0] in {"agents"}: initial_load_tokens += tokens + if top_dir in DEFERRED_RESOURCE_DIRS: + deferred_resource_tokens += tokens + current = deferred_resource_dirs.setdefault( + top_dir, + {"path": top_dir, "estimated_tokens": 0, "file_count": 0}, + ) + current["estimated_tokens"] = int(current["estimated_tokens"]) + tokens + current["file_count"] = int(current["file_count"]) + 1 budget_tier = budget_tier_for(manifest) budget_limit = max_initial_tokens if max_initial_tokens is not None else CONTEXT_BUDGETS[budget_tier] @@ -174,6 +188,22 @@ def analyze_skill( if other_tokens and skill_body_tokens / (skill_body_tokens + other_tokens) > 0.75: warnings.append("Most text still lives in SKILL.md; consider moving detail into references/ or scripts/.") + large_deferred_resource_dirs = [ + item + for item in sorted( + deferred_resource_dirs.values(), + key=lambda payload: int(payload["estimated_tokens"]), + reverse=True, + ) + if int(item["estimated_tokens"]) > DEFERRED_RESOURCE_WARN_DIR_TOKENS + ] + if deferred_resource_tokens > DEFERRED_RESOURCE_WARN_TOKENS: + warnings.append( + "Deferred resource footprint is high: " + f"{deferred_resource_tokens} estimated tokens across references/scripts/evals. " + "Keep Review Studio warnings visible until the largest resource dirs are split, archived, or justified." + ) + frontmatter = read_frontmatter(skill_md) governance_score, _ = compute_score(root, manifest, frontmatter, skill_text, bool(manifest)) signal_points = quality_signal_points(root, manifest, governance_score) @@ -190,6 +220,14 @@ def analyze_skill( "other_text_tokens": other_tokens, "estimated_initial_load_tokens": initial_load_tokens, "estimated_total_text_tokens": total_text_tokens, + "deferred_resource_tokens": deferred_resource_tokens, + "deferred_resource_warn_threshold": DEFERRED_RESOURCE_WARN_TOKENS, + "deferred_resource_dirs": sorted( + deferred_resource_dirs.values(), + key=lambda payload: int(payload["estimated_tokens"]), + reverse=True, + ), + "large_deferred_resource_dirs": large_deferred_resource_dirs, "relevant_file_count": len(files), "unused_resource_dirs": unused_resource_dirs, "quality_signal_points": signal_points, diff --git a/scripts/review_studio_gates.py b/scripts/review_studio_gates.py index 9dc2a33..584a9ac 100644 --- a/scripts/review_studio_gates.py +++ b/scripts/review_studio_gates.py @@ -200,9 +200,18 @@ def build_gates(skill_dir: Path, output_html: Path, data: dict[str, dict[str, An context_status = "warn" if context_status == "pass" else context_status if not context: context_status = "warn" + large_deferred_dirs = context_stats.get("large_deferred_resource_dirs", []) or [] + top_deferred = "none" + if large_deferred_dirs: + first = large_deferred_dirs[0] + top_deferred = f"{first.get('path', 'resource')} {first.get('estimated_tokens', 'n/a')}" context_detail = ( f"initial load {context_stats.get('estimated_initial_load_tokens', 'n/a')}/" - f"{context_stats.get('context_budget_limit', 'n/a')}; quality density {context_stats.get('quality_density', 'n/a')}" + f"{context_stats.get('context_budget_limit', 'n/a')}; " + f"deferred {context_stats.get('deferred_resource_tokens', 'n/a')}/" + f"{context_stats.get('deferred_resource_warn_threshold', 'n/a')}; " + f"top deferred {top_deferred}; " + f"quality density {context_stats.get('quality_density', 'n/a')}" ) gates.append( gate( diff --git a/tests/verify_quality_checks.py b/tests/verify_quality_checks.py index 1954a59..e94502d 100644 --- a/tests/verify_quality_checks.py +++ b/tests/verify_quality_checks.py @@ -44,16 +44,25 @@ def require_context_targets(case: dict, max_initial: int, min_density: float) -> stats = case.get("payload", {}).get("stats", {}) initial = stats.get("estimated_initial_load_tokens") density = stats.get("quality_density") + deferred = stats.get("deferred_resource_tokens") + deferred_threshold = stats.get("deferred_resource_warn_threshold") + deferred_dirs = stats.get("deferred_resource_dirs", []) unused = stats.get("unused_resource_dirs", []) case["max_initial_load"] = max_initial case["observed_initial_load"] = initial case["minimum_quality_density"] = min_density case["observed_quality_density"] = density + case["observed_deferred_resource_tokens"] = deferred + case["deferred_resource_warn_threshold"] = deferred_threshold + case["deferred_resource_dir_count"] = len(deferred_dirs) if isinstance(deferred_dirs, list) else None case["unused_resource_dirs"] = unused case["passed"] = ( case["passed"] and initial is not None and density is not None + and deferred is not None + and deferred_threshold is not None + and isinstance(deferred_dirs, list) and initial <= max_initial and density >= min_density and not unused diff --git a/tests/verify_review_studio.py b/tests/verify_review_studio.py index dc0bc51..28658ab 100644 --- a/tests/verify_review_studio.py +++ b/tests/verify_review_studio.py @@ -116,6 +116,13 @@ def main() -> None: capture_output=True, text=True, ) + subprocess.run( + [sys.executable, str(ROOT / "scripts" / "render_context_reports.py")], + cwd=ROOT, + check=True, + capture_output=True, + text=True, + ) subprocess.run( [ sys.executable, @@ -279,20 +286,21 @@ def main() -> None: assert payload["schema_version"] == "2.0", payload assert payload["summary"]["decision"] == "review", payload assert payload["summary"]["gate_count"] == 16, payload - assert payload["summary"]["world_class_score"] == 91, payload - assert payload["summary"]["warning_count"] == 3, payload + assert payload["summary"]["world_class_score"] == 89, payload + assert payload["summary"]["warning_count"] == 4, payload assert payload["summary"]["blocker_count"] == 0, payload - assert payload["summary"]["action_count"] == 3, payload + assert payload["summary"]["action_count"] == 4, payload assert payload["summary"]["annotation_count"] == 0, payload assert payload["summary"]["open_annotation_blocker_count"] == 0, payload assert payload["summary"]["action_count"] == payload["summary"]["warning_count"] + payload["summary"]["blocker_count"], payload assert {item["gate_key"] for item in payload["review_actions"]} == { "output-lab", + "context-budget", "review-waivers", "world-class-evidence", }, payload gate_keys = {item["key"] for item in payload["gates"]} - assert {"intent-canvas", "trigger-lab", "output-lab", "runtime-matrix", "trust-report", "python-compat", "architecture-maintainability", "permission-gates", "permission-runtime", "skill-atlas", "operations-loop", "review-waivers", "world-class-evidence", "registry-audit", "release-notes"} <= gate_keys, payload + assert {"intent-canvas", "trigger-lab", "output-lab", "context-budget", "runtime-matrix", "trust-report", "python-compat", "architecture-maintainability", "permission-gates", "permission-runtime", "skill-atlas", "operations-loop", "review-waivers", "world-class-evidence", "registry-audit", "release-notes"} <= gate_keys, payload output_gate = next(item for item in payload["gates"] if item["key"] == "output-lab") assert output_gate["status"] == "warn", output_gate assert "5/5 cases" in output_gate["detail"], output_gate @@ -302,6 +310,13 @@ def main() -> None: assert "model 0" in output_gate["detail"], output_gate assert "reviewed 0/5" in output_gate["detail"], output_gate assert "review pending 5" in output_gate["detail"], output_gate + context_gate = next(item for item in payload["gates"] if item["key"] == "context-budget") + assert context_gate["status"] == "warn", context_gate + assert "initial load 944/1000" in context_gate["detail"], context_gate + assert "deferred " in context_gate["detail"], context_gate + assert "/120000" in context_gate["detail"], context_gate + assert "top deferred scripts" in context_gate["detail"], context_gate + assert "quality density" in context_gate["detail"], context_gate release_gate = next(item for item in payload["gates"] if item["key"] == "release-notes") assert "upgrade minor declared / minor recommended" in release_gate["detail"], release_gate assert "reports/upgrade_check.json" in release_gate["evidence"], release_gate @@ -347,7 +362,7 @@ def main() -> None: assert "reports/adoption_drift_report.json" in operations_gate["evidence"], operations_gate waivers_gate = next(item for item in payload["gates"] if item["key"] == "review-waivers") assert waivers_gate["status"] == "warn", waivers_gate - assert "1 warning gates still need reviewer decision" in waivers_gate["detail"], waivers_gate + assert "2 warning gates still need reviewer decision" in waivers_gate["detail"], waivers_gate assert "reports/review_waivers.json" in waivers_gate["evidence"], waivers_gate world_class_gate = next(item for item in payload["gates"] if item["key"] == "world-class-evidence") assert world_class_gate["status"] == "warn", world_class_gate @@ -411,7 +426,17 @@ def main() -> None: assert full_payload["data"]["architecture_maintainability"]["summary"]["hotspot_count"] == 0, full_payload["data"]["architecture_maintainability"] assert full_payload["data"]["architecture_maintainability"]["summary"]["blocker_count"] == 0, full_payload["data"]["architecture_maintainability"] action_keys = {item["gate_key"] for item in full_payload["review_actions"]} - assert action_keys == {"output-lab", "review-waivers", "world-class-evidence"}, full_payload["review_actions"] + assert action_keys == {"output-lab", "context-budget", "review-waivers", "world-class-evidence"}, full_payload["review_actions"] + context_action = next(item for item in full_payload["review_actions"] if item["gate_key"] == "context-budget") + assert "deferred resources" in context_action["summary"], context_action + assert {item["path"] for item in context_action["source_refs"]} >= { + "SKILL.md", + "reports/context_budget.md", + "scripts/resource_boundary_check.py", + "references/skill-engineering-method.md", + }, context_action + assert all(item["exists"] for item in context_action["source_refs"]), context_action + assert "python3 scripts/render_context_reports.py" == context_action["verification_command"], context_action world_class_action = next(item for item in full_payload["review_actions"] if item["gate_key"] == "world-class-evidence") assert {item["path"] for item in world_class_action["source_refs"]} >= { "reports/world_class_evidence_ledger.md", @@ -505,6 +530,7 @@ def main() -> None: assert "审查闸门" in html, html[:1200] assert "修复动作" in html, html[:3000] assert "补足 output eval 覆盖、execution evidence、blind A/B 和 reviewer adjudication。" in html, html[:9000] + assert "压缩或拆分高成本 deferred resources" in html, html[:9000] assert "reports/output_review_kit.html" in html, html[:9000] assert "python3 scripts/adjudicate_output_review.py --write-template" in html, html[:9000] assert "对保留的 warning 写入 reviewer、理由、范围和到期时间，或修掉 warning。" in html, html[:9000]