最近功能文档覆盖矩阵¶
本页用于回答一个运维读者最常见的问题：changelog 里提到的新能力，我应该去哪里理解、配置、验证？
覆盖口径
本矩阵只记录已在当前实现中存在的功能与文档入口。v0.6.1 后配置入口以 CS_* / AHP_*、clawsentry start、clawsentry config show --effective 和显式 --env-file 为准；.clawsentry.env.example 只是可提交模板，本地 env 文件只作为显式 runtime input 或 legacy migration surface。
覆盖矩阵¶
| 功能 / 能力 | 源码或验证证据 | Web UI / API / CLI 入口 | 主要文档 | 状态 |
|---|---|---|---|---|
| Default scope profile and Web UI clarity | src/clawsentry/gateway/server.py；src/clawsentry/gateway/env_config.py；src/clawsentry/ui/src/pages/SessionDetail.tsx；src/clawsentry/tests/test_gateway.py；docs/validation/v0.6.6-default-scope-webui-release-2026-05-02.md |
CS_SESSION_SCOPE_PROFILE_FILE；Gateway pre_action decisions；Dashboard Session Detail risk charts；Web UI login |
Session scope 配置、环境变量、Dashboard、Sanitizer 当前能力 | v0.6.6 released |
| Scope preview and sanitizer clarity | src/clawsentry/cli/scope_command.py；src/clawsentry/gateway/session_scope.py；src/clawsentry/tests/test_scope_command.py；docs/validation/v0.6.5-scope-preview-sanitizer-release-2026-05-02.md |
clawsentry scope validate/preview；POST /ahp/scope/preview；clawsentry watch |
Session scope 配置、Sanitizer 当前能力、Decision API scope preview、CLI 命令 | v0.6.5 released |
| Env-file discovery hints | src/clawsentry/cli/dotenv_loader.py；src/clawsentry/tests/test_dotenv_loader.py；src/clawsentry/tests/test_start_command.py；src/clawsentry/tests/test_config_command.py；docs/validation/v0.6.4-env-file-hints-release-2026-04-30.md |
clawsentry start --framework <name>；clawsentry start --env-file .clawsentry.env.local；clawsentry config show --effective --env-file <path> |
配置概览、配置模板、环境变量、CLI 命令 | v0.6.4 release prep |
| Env-first config docs refresh | docs/validation/v0.6.3-env-first-config-docs-release-2026-04-30.md；src/clawsentry/tests/test_public_docs_contract.py；src/clawsentry/tests/test_ux_config_docs_benchmark_contract.py |
clawsentry config show --effective；clawsentry config wizard；显式 --env-file / CLAWSENTRY_ENV_FILE |
配置概览、环境变量、配置模板、检测管线配置、策略调优 | v0.6.3 release |
| Strict config-source split + explicit env-file | docs/validation/v0.6.1-config-source-redesign-release-2026-04-29.md；src/clawsentry/tests/test_project_config.py；src/clawsentry/tests/test_dotenv_loader.py；src/clawsentry/tests/test_start_command.py |
CS_FRAMEWORK / CS_ENABLED_FRAMEWORKS；clawsentry start --env-file <path>；CLAWSENTRY_ENV_FILE=<path>；clawsentry config show --effective --env-file <path> |
配置概览、环境变量、配置模板、快速开始 | v0.6.1 release |
| Setup/config precedence + L3 routing E2E | docs/validation/v0.6.0-setup-docs-l3-e2e-release-2026-04-29.md；docs/validation/2026-04-28-clawsentry-config-truth-matrix.md；src/clawsentry/tests/test_l3_real_provider_e2e.py |
clawsentry config wizard --interactive；CS_L3_ROUTING_MODE=replace_l2；CS_L3_TRIGGER_PROFILE=eager；/report/session/{id}/risk L3 summary |
配置概览、配置模板、检测管线配置、Reporting API | v0.6.0 release-ready |
| Anti-bypass Follow-up Guard | src/clawsentry/gateway/anti_bypass_guard.py；src/clawsentry/tests/test_anti_bypass_guard.py；docs/validation/v0.5.14-anti-bypass-release-2026-04-28.md |
CS_ANTI_BYPASS_GUARD_ENABLED=true；decision SSE redacted anti_bypass metadata；defer-pending SSE redacts retry command to tool name；Gateway decision path |
Anti-bypass Guard 决策引擎、检测管线配置、环境变量 | released |
| L3 advisory full-review / snapshot / job / review | src/clawsentry/gateway/server.py 的 /report/*/l3-advisory/* 路由；docs/validation/v0.5.10-webui-l3-ux-release-2026-04-26.md |
Session Detail 的 Request L3 full review；clawsentry l3 full-review；POST /report/session/{session_id}/l3-advisory/full-review |
L3 咨询审查、Reporting API | covered |
| Bounded L3 job drain | GET /report/l3-advisory/jobs、POST /report/l3-advisory/jobs/run-next、POST /report/l3-advisory/jobs/drain；validation release evidence |
clawsentry l3 jobs list/run-next/drain；API jobs endpoints |
L3 咨询审查：queued jobs、API validity report | covered |
| Heartbeat-compatible aggregate queueing | trigger_reason=heartbeat_aggregate 验证；heartbeat / idle / success / rate_limit 事件兼容聚合路径；validation release evidence |
Feature gates CS_L3_ADVISORY_ASYNC_ENABLED=true + CS_L3_HEARTBEAT_REVIEW_ENABLED=true；冻结 snapshot 并入队 job，不自动运行 |
L3 咨询审查：heartbeat / idle aggregate queueing | covered |
| L3 advisory natural-language action payload | docs/validation/v0.5.10-webui-l3-ux-release-2026-04-26.md；review/action payload 字段 analysis_summary、analysis_points、operator_next_steps |
Session Detail 的 Analysis summary / points / next steps；SSE/watch/report action payload | L3 咨询审查：Web UI | covered |
| 同步 L3 Agent 触发与可观测字段 | src/clawsentry/gateway/l3_trigger.py；l3_state、trigger_reason、trigger_detail、budget_exhaustion_event report fields |
clawsentry watch、Runtime Feed、Session Detail、/report/session/{id}/risk |
L3 审查 Agent、L2 语义分析 | refreshed |
| Kimi CLI native hooks | src/clawsentry/adapters/kimi_adapter.py; src/clawsentry/tests/test_kimi_*; docs/validation/kimi-cli-real-hook-feasibility-2026-04-29.md |
clawsentry init kimi-cli --setup; clawsentry harness --framework kimi-cli; $KIMI_SHARE_DIR/config.toml / ~/.kimi/config.toml |
Kimi CLI 集成 | native-hook support; no modify/defer parity |
| Gemini CLI native hooks | src/clawsentry/adapters/gemini_adapter.py；docs/validation/gemini-cli-real-hook-feasibility-2026-04-25.md |
clawsentry init gemini-cli --setup；clawsentry harness --framework gemini-cli；project .gemini/settings.json |
Gemini CLI 集成、首页 Gemini path | covered |
| Benchmark mode | clawsentry benchmark env|enable|disable|run；docs/validation/v0.5.9-docs-runtime-webui-release-2026-04-26.md |
CS_MODE=benchmark；Codex benchmark uses temp CODEX_HOME |
Benchmark 模式、配置模板：CI / benchmark | covered |
| Metric / window fields | src/clawsentry/gateway/session_registry.py；docs/validation/v0.5.12-metric-wizard-agentdog-progress-2026-04-27.md；Dashboard Sessions / Session Detail fields |
/report/sessions、/report/session/{id}/risk、SSE、Dashboard cards |
指标字典、Reporting API、Dashboard | release-ready |
| Web UI L3 surfaces | src/clawsentry/ui/src/pages/SessionDetail.tsx；docs/validation/v0.5.10-webui-l3-ux-release-2026-04-26.md |
Runtime Feed、Sessions、Session Detail L3 advisory card/action | Dashboard、L3 咨询审查 | covered |
| Token budget / LLM usage | InstrumentedProvider usage fields；CS_LLM_TOKEN_BUDGET_*; v0.5.10 token-first UI notes |
LLM drilldown/status, budget SSE event, config show --effective |
LLM 配置、环境变量 | refreshed |
| Interactive config wizard | src/clawsentry/cli/config_command.py TTY prompt flow；docs/validation/v0.5.12-metric-wizard-agentdog-progress-2026-04-27.md |
clawsentry config wizard --interactive；clawsentry config wizard --non-interactive ... |
快速开始、配置模板、CLI 命令 | release-ready |
| Multi-framework startup | clawsentry start --framework/--frameworks readiness summaries; v0.4.x/v0.5.x changelog |
clawsentry integrations status、clawsentry start --framework codex |
快速开始、CLI 命令、各集成页 | covered |
| Latch integration | clawsentry start --with-latch、clawsentry latch status/start/install |
Hub UI / Gateway start banner / troubleshooting | Latch 集成、故障排查：Latch | covered |
| OpenClaw managed setup boundary | clawsentry init openclaw 默认只打印 env 建议；--setup / --setup-openclaw explicit opt-in |
OpenClaw integration status and webhook/WS paths | OpenClaw 集成、CLI 命令 | covered |
| Codex managed setup boundary | Managed native hooks and CODEX_NATIVE_HOOKS doctor details; validation evidence under release evidence docs |
clawsentry init codex --setup、clawsentry doctor、temporary CODEX_HOME for real hook tests |
Codex CLI 集成、故障排查：Codex Session Watcher | covered |
Docs audit checklist¶
- [x] L2/L3 pages explicitly distinguish L1 deterministic rules, L2 semantic analysis, synchronous L3 Agent, and L3 advisory full-review.
- [x] L3 advisory docs state latency/budget/runner boundaries, queued-only drain semantics, heartbeat-compatible aggregate gates, and
advisory_only=true/canonical_decision_mutated=false. - [x] LLM config docs state that L3 advisory
llm_providerinherits sharedCS_LLM_*settings and degrades loudly when they are missing. - [x] Recent Gemini CLI, benchmark, Latch, OpenClaw and Codex setup boundaries have primary pages and validation evidence links.
- [x] Metric/window field deep rewrite is complete in the metric dictionary lane; this matrix links to the primary reference instead of duplicating field semantics.
- [x] Interactive config wizard copy now distinguishes project config writing from framework hook installation.
Follow-up watchlist¶
| Follow-up | Reason | Owner lane |
|---|---|---|
| AgentDoG labeled ATBench sample set | Smoke replay is complete, but scored safety metrics require labeled safe/unsafe records | Benchmark lane |
| Raw vs ClawSentry live runners | Offline replay proves detection infrastructure, not live framework prevention | Benchmark lane |
Optional clawsentry setup alias |
Tested setup surface is config wizard; add a shorter alias only if a release needs it |
Separate optional CLI lane |