@zenalexa/unicli 0.217.0 → 0.218.0

package/AGENTS.md

@@ -1,215 +1,55 @@
 # Uni-CLI
 
-<!-- IF YOU ARE AN AI AGENT: Read this file. It tells you everything you need.
-     Install: npm install -g @zenalexa/unicli
-     Discover: unicli search "your intent"
-     Agent routing: unicli agents matrix && unicli agents recommend <agent>
-     Run: unicli <site> <command> [args]
-     Output is structured Markdown by default (non-TTY + agent UA auto-detected). Use -f json for JSON. Errors are structured envelopes to stderr.
-     You can fix broken adapters yourself — read the Self-Repair section. -->
+> AI agents: install with `npm install -g @zenalexa/unicli`. Discover with
+> `unicli search "intent"`. Run with `unicli <site> <command>`. Output is
+> Markdown when piped, JSON via `-f json`. Errors are structured envelopes
+> on stderr. An `auth_required` envelope means the site needs cookies —
+> run `unicli auth setup <site>` once and retry. Adapter is broken?
+> Read `unicli repair <site> <command>`.
 
-<!-- BEGIN COUNTS -->
+Universal CLI for websites, desktop apps, and local tools. YAML adapters are
+self-repairing — agents can fix them by editing one file. See
+`docs/BENCHMARK.md` for measured per-call token costs.
 
-> <!-- STATS:site_count -->235<!-- /STATS --> sites, <!-- STATS:command_count -->1448<!-- /STATS --> commands, <!-- STATS:pipeline_step_count -->59<!-- /STATS --> pipeline steps, BM25 bilingual search. `npm install -g @zenalexa/unicli`
-
-<!-- END COUNTS -->
-
-## For AI Agents
-
-Universal CLI for websites, desktop apps, and local tools. Markdown when piped (structured envelope). Self-repairing YAML adapters.
-
-```bash
-unicli search "twitter trending"    # Find commands by intent
-unicli <site> <command> [options]    # Run any command
-unicli repair <site> <command>       # Diagnose + fix a broken adapter
-unicli list                          # All commands (MD when piped)
-```
-
-## Install
-
-```bash
-npm install -g @zenalexa/unicli
-```
-
-<!-- BEGIN ADAPTERS -->
-
-## What You Can Do
-
-### Web (138+ sites)
-
-**Chinese**: zhihu (24), xiaohongshu (22), bilibili (20), douyin (13), douban (12), v2ex (12), linux-do (11), jike (10), +31 more (`unicli list`)
-
-**International**: twitter (38), instagram (26), reddit (20), tiktok (16), youtube (16), nowcoder (16), discord-app (15), lesswrong (15), +48 more (`unicli list`)
-
-**AI / ML**: antigravity (16), chatwise (16), chatgpt (15), notebooklm (15), doubao-app (13), doubao (9), doubao-web (9), perplexity (8), +13 more (`unicli list`)
-
-**Finance**: eastmoney (18), xueqiu (14), binance (13), sinafinance (5), barchart (4), yahoo-finance (3), coinbase (2), futu (2)
-
-**Developer**: cursor (18), codex (17), vscode (10), docker-desktop (7), github-desktop (7), gitkraken (7), insomnia (7), postman (7), +16 more (`unicli list`)
-
-**News**: bloomberg (10), hackernews (10), 36kr (5), bbc (4), reuters (4), ithome (3), cnn (2), infoq (2), +3 more (`unicli list`)
-
-**Reference**: spotify (23), netease-music (17), linear (10), imdb (7), bitwarden (7), todoist (7), wikipedia (5), xiaoyuzhou (5), +11 more (`unicli list`)
-
-### macOS (58 cmds)
-
-active-app, apps, apps-list, battery, bluetooth, brightness, caffeinate, calendar-create, calendar-list, calendar-today, clipboard, contacts-search, … (`unicli list --site macos`)
-
-### Desktop (28 apps)
-
-freecad (15 cmds), blender (13 cmds), gimp (12 cmds), ffmpeg (11 cmds), audacity (8 cmds), figma (8 cmds), docker (7 cmds), excel (7 cmds), +20 more (`unicli list --category desktop`)
-
-### Bridge (3 CLIs)
-
-gh (6 cmds), jq (2 cmds), yt-dlp (4 cmds)
-
-<!-- END ADAPTERS -->
-
-## Authentication
-
-Some sites require cookies:
-
-```bash
-unicli auth setup <site>    # Show required cookies + template
-unicli auth check <site>    # Validate cookie file
-unicli auth list            # List configured sites
-```
-
-Cookie file format: `{ "SESSDATA": "value", "bili_jct": "value" }`. Store at `~/.unicli/cookies/<site>.json`.
-
-Sites requiring auth: bilibili, weibo, zhihu, twitter, xueqiu, zsxq, jike, weread, douban, linux-do, v2ex (some commands).
-
-## Output Contract
-
-Commands return v2 `AgentEnvelope` on stdout across adapter, core, ext/dev, and admin surfaces (`agents matrix/recommend/generate`, `auth`, `mcp`, `repair`, etc.). `mcp serve` and `acp` stay raw stdio protocol servers. Format auto-selected — pipe or set an agent UA env var for Markdown.
-
-### Format auto-selection
-
-Priority: `-f` flag > `UNICLI_OUTPUT` env > non-TTY or agent-UA env (`md`) > `md` default. Values: `json | yaml | md | csv | compact`. Agent-UA env vars: `CLAUDE_CODE`, `CODEX_CLI`, `OPENCODE`, `HERMES_AGENT`, `UNICLI_AGENT`.
-
-### Envelope shape
-
-Success:
-
-```yaml
-ok: true
-schema_version: "2"
-command: "twitter.mentions"
-meta:
-  duration_ms: 412
-  count: 20
-  surface: web # web | desktop | system | mobile
-  pagination:
-    has_more: true
-    next_cursor: "abc123"
-data:
-  - { id: "...", text: "...", author: "..." }
-error: null
-```
-
-Error:
-
-```yaml
-ok: false
-schema_version: "2"
-command: "twitter.mentions"
-meta:
-  duration_ms: 91
-data: null
-error:
-  code: auth_required # see error codes below
-  message: "401 Unauthorized"
-  adapter_path: "src/adapters/twitter/mentions.yaml"
-  step: 1
-  suggestion: "Run: unicli auth setup twitter"
-  retryable: false
-  alternatives: ["twitter.search", "twitter.timeline"]
-```
-
-### MD body sections
-
-Success: `## Data` (per-item list) · `## Context` (surface, pagination) · `## Next Actions`. Error: `## Error` (code, message, adapter_path, step, retryable) · `## Suggestion` · `## Alternatives`. YAML frontmatter carries `ok`, `schema_version`, `command`, `duration_ms`, `count`.
-
-### Error codes
-
-net: `network_error` `rate_limited` `upstream_error` `api_error` `not_authenticated`
-input: `invalid_input` `selector_miss` `not_found`
-authz: `auth_required` `permission_denied`
-runtime: `internal_error` `quarantined`
-ref: `stale_ref` `ambiguous` `ref_not_found`
-
-### Exit codes
-
-0 ok · 66 empty · 69 unavailable · 75 temp-fail · 77 auth · 78 config
-
-## Self-Repair Protocol
-
-When a command fails:
+## Done = these commands exit 0
 
 ```
-1. Read error envelope (MD or JSON) → get adapter_path
-2. Open the YAML (~20 lines, no imports)
-3. Edit the failing step (URL changed, selector moved, auth needed)
-4. Save to ~/.unicli/adapters/<site>/<command>.yaml
-5. Verify: unicli repair <site> <command>
+pnpm typecheck && pnpm lint && pnpm test:unit
 ```
 
-Fixes persist in `~/.unicli/adapters/` and survive `npm update`.
-
-## Creating Adapters
+Full E2E + adapter coverage: `npm run verify`. Required before any release.
 
-```yaml
-site: example
-name: search
-type: web-api
-strategy: public
-pipeline:
-  - fetch: { url: "https://api.example.com/search?q=${{ args.query }}" }
-  - select: data.results
-  - map: { title: "${{ item.title }}", url: "${{ item.url }}" }
-  - limit: ${{ args.limit }}
-args:
-  query: { type: str, required: true, positional: true }
-  limit: { type: int, default: 20 }
-columns: [title, url]
-```
+## Project conventions
 
-Full reference: [`docs/ADAPTER-FORMAT.md`](docs/ADAPTER-FORMAT.md).
+The cross-CLI contract `~/.claude/AGENTS.md` and ruleset `~/.claude/rules/`
+apply in full. The bullets below are the project-specific reinforcement
+because Uni-CLI is adapter-heavy and patch-rot kills us faster than most
+codebases.
 
-## Pipeline Steps
+- **Engine code lives in `src/engine/`, browser in `src/browser/`, commands in `src/commands/`, adapters in `src/adapters/`.** Map by responsibility — never by version.
+- **Errors emit structured envelopes** to stderr with `code`, `adapter_path`, `step`, `suggestion`. Pipeline steps that fail must surface the real cause, never coerce to a generic `internal_error`. (rule 02)
+- **Tests under `tests/` and `*.test.ts` exercise real owned code** — engine, registry, adapter loader. External boundaries (network fetch, subprocess, Chrome CDP) may be stubbed with one `// REASON:` line. (rule 03)
+- **`unicli test [site]` runs adapter E2E.** Never substitute a fixture for the YAML pipeline runner.
+- **Multi-file change in `src/engine/`, `src/browser/`, or new adapter type → independent audit subagent before PR.** (rule 05)
 
-59 registered pipeline steps across api, transform, control, browser, subprocess, and CUA-oriented families. Common steps include fetch, fetch_text, parse_rss, html_to_md, select, map, filter, sort, limit, set, if, each, parallel, rate_limit, assert, retry, append, navigate, evaluate, click, type, press, scroll, intercept, snapshot, tap, extract, wait, exec, write_temp, download, and websocket. See [`docs/ADAPTER-FORMAT.md`](docs/ADAPTER-FORMAT.md) for examples.
+## Style template
 
-## MCP Server
+Detailed adapter format, pipeline conventions, and error envelope shape live
+in the project skill: `~/.claude/skills/uni-cli-style/SKILL.md`. Load on
+demand. Never paraphrase from memory.
 
-```bash
-npx @zenalexa/unicli mcp serve                                # stdio (4 meta-tools)
-npx @zenalexa/unicli mcp serve --transport streamable --port 19826
-npx @zenalexa/unicli mcp serve --transport sse --port 19826   # SSE for remote
-npx @zenalexa/unicli mcp serve --auth                         # OAuth 2.1 PKCE
-```
+## Project-internal references
 
-Default tools: `unicli_run`, `unicli_list`, `unicli_search`, `unicli_explore`. Agents call `unicli_search` first (bilingual BM25), then `unicli_run` with the chosen site + command.
+| Topic                       | Where                    |
+| --------------------------- | ------------------------ |
+| Adapter format              | `docs/ADAPTER-FORMAT.md` |
+| Pipeline steps (live count) | `unicli list`            |
+| Strategy semantics          | `src/types.ts`           |
+| Theory / citations          | `docs/THEORY.md`         |
 
-## ACP (avante.nvim / Zed)
-
-```bash
-unicli acp          # ACP compatibility over stdio
-```
-
-ACP is an editor compatibility gateway, not the core runtime. Prefer `unicli agents recommend <agent>` for Claude Code, Codex, Cursor, OpenCode, Gemini, Qwen, Kiro, Aider, Goose, Cline, Roo, AgentAPI, etc.
-
-## External CLI Passthrough
-
-58 registered external CLIs. If installed on your system, they're available as top-level commands:
-
-```bash
-unicli ext list                    # Show all external CLIs + install status
-unicli ext list --tag agent        # Show coding-agent CLIs
-unicli ext install <name>          # Install an external CLI
-unicli lark-cli calendar +agenda   # Direct passthrough
-```
+`unicli list` is more authoritative than any inventory in this file — the
+project ships at high cadence, written counts go stale fast.
 
 ## Version
 
-0.217.0 — Apollo · Lovell
+0.218.0 — Apollo · Cernan

package/README.md

@@ -28,7 +28,7 @@
 </p>
 
 <p align="center">
-  <sub><!-- STATS:site_count -->235<!-- /STATS --> sites · <!-- STATS:command_count -->1448<!-- /STATS --> commands · <!-- STATS:pipeline_step_count -->59<!-- /STATS --> pipeline steps · <!-- STATS:test_count -->7471<!-- /STATS --> tests</sub>
+  <sub><!-- STATS:site_count -->237<!-- /STATS --> sites · <!-- STATS:command_count -->1453<!-- /STATS --> commands · <!-- STATS:pipeline_step_count -->101<!-- /STATS --> pipeline steps · <!-- STATS:test_count -->7842<!-- /STATS --> tests</sub>
 </p>
 
 <p align="center">
@@ -131,26 +131,28 @@ Prefer native CLI / JSON stream / MCP for agent runtimes. Use ACP as an editor c
 
 Uni-CLI sits under agent applications and turns software surfaces into commands that agents can discover, execute, record, and repair.
 
-| Surface            | What you get                                                                                       |
-| ------------------ | -------------------------------------------------------------------------------------------------- |
-| Websites and APIs  | Declarative adapters for public, cookie, header, and browser-intercept workflows                   |
-| Browser automation | CDP steps for navigate, click, type, intercept, snapshot, extract, wait, and related browser work  |
-| Desktop and macOS  | System commands, app adapters, screenshots, clipboard, calendar, brightness, and local tools       |
-| External CLIs      | 58 registered pass-through bridges with install/status discovery                                   |
-| Agent backends     | Route matrix for native CLI, JSON stream, MCP, ACP, HTTP API, OpenAI-compatible, and bridge routes |
-| Operation policy   | `open`, `confirm`, and `locked` profiles with effect, risk, approval, and capability scope         |
-| Evidence           | Optional run traces plus browser pre/post evidence, movement detection, stale-ref details          |
-| Output             | v2 `AgentEnvelope` in Markdown, JSON, YAML, CSV, or compact format                                 |
-| Repair             | Structured errors with `adapter_path`, failing `step`, retryability, suggestions, and alternatives |
+| Surface            | What you get                                                                                                                                                                                             |
+| ------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| Websites and APIs  | Declarative adapters for public, cookie, header, and browser-intercept workflows                                                                                                                         |
+| Browser automation | CDP steps for navigate, click, type, intercept, snapshot, extract, wait, and related browser work                                                                                                        |
+| Desktop and macOS  | System commands, app adapters, real-time Shortcuts/App Intent discovery, screenshots, clipboard, calendar, brightness, and local tools                                                                   |
+| External CLIs      | 58 registered pass-through bridges with install/status discovery                                                                                                                                         |
+| Agent backends     | Route matrix for native CLI, JSON stream, MCP, ACP, HTTP API, OpenAI-compatible, and bridge routes                                                                                                       |
+| Operation policy   | `open`, `confirm`, and `locked` profiles with effect/risk scopes, local deny rules, `--yes`, and persisted approval memory                                                                               |
+| Evidence           | Run traces with environment snapshots, probe/replay/compare scores, structured gate results, browser session leases with tab/auth posture, render-aware evidence, movement checks, and stale-ref details |
+| Output             | v2 `AgentEnvelope` in Markdown, JSON, YAML, CSV, or compact format                                                                                                                                       |
+| Repair             | Structured errors with `adapter_path`, failing `step`, retryability, suggestions, and alternatives                                                                                                       |
 
 ## For Agents
 
 Use search first, then run the smallest matching command.
 
 ```bash
-unicli search "twitter trending" --limit 5
-unicli twitter search "coding agents" -f json
-unicli repair twitter search
+unicli search "connect slack messages" --limit 5
+unicli slack search "deploy incident" -f json
+unicli macos app-actions --app WhatsApp -f json
+unicli macos automation-smoke -f json
+unicli repair slack search
 ```
 
 Output defaults to structured Markdown for non-TTY and agent-user-agent runs. Force a machine format when you need one:
@@ -159,6 +161,16 @@ Output defaults to structured Markdown for non-TTY and agent-user-agent runs. Fo
 UNICLI_OUTPUT=json unicli reddit hot --limit 10
 unicli hackernews top --limit 5 -f yaml
 unicli --record --permission-profile confirm twitter search "coding agents" -f json
+unicli runs list -f json
+unicli runs show <run_id> -f json
+unicli runs probe <run_id> -f json
+unicli runs replay <run_id> --permission-profile confirm --yes --min-score 1 --min-context-score 1 --min-overall-score 1 -f json
+unicli runs compare <run_id> <replay_run_id> -f json
+unicli runs compare <run_id> <replay_run_id> --min-score 1 --min-context-score 1 --min-overall-score 1 -f json
+unicli --permission-profile locked --yes --remember-approval word set-font "Inter"
+unicli approvals list -f json
+unicli approvals revoke <approval_key> -f json
+unicli browser evidence --render-aware --expect-domain example.com -f json
 ```
 
 Protocol entry points:
@@ -173,9 +185,25 @@ unicli agents matrix
 
 ACP is supported for editors and bridge tooling. The primary runtime path stays native CLI, JSON stream, or MCP when those routes are available.
 
+## Local Computer Control
+
+`unicli compute` controls installed apps through native accessibility, Electron CDP, and visual fallback transports.
+
+```bash
+unicli compute apps
+unicli compute snapshot --app Calculator --format compact
+unicli compute find --role button --name 5 --first
+unicli compute find --role input --text 8 --first
+unicli compute click @e7
+unicli doctor compute --json
+npx @zenalexa/unicli mcp serve --profile computer-use
+```
+
+Start with [Compute](docs/operate/compute.md), [Electron App Control](docs/operate/electron.md), and [Compute Troubleshooting](docs/operate/troubleshooting.md).
+
 ## Coverage
 
-The catalog is intentionally broad, but the important point is not the count. The important point is that every command is discoverable, typed, and repairable.
+The catalog is intentionally broad. Every command is discoverable, typed, and repairable.
 
 | Area                    | Examples                                                                                                    |
 | ----------------------- | ----------------------------------------------------------------------------------------------------------- |
@@ -293,12 +321,24 @@ Docs:
 - Browser adapters require a reachable Chrome/CDP session.
 - Permission profiles are user-selected runtime policy. The default is `open`;
   stricter `confirm` and `locked` profiles require `--yes` or `UNICLI_APPROVE=1`
-  for blocked operations.
+  for blocked operations. Add `--remember-approval` with `--yes` to store the
+  same command capability and resource scope under `~/.unicli/approvals.jsonl`.
+  Resource scope covers stable metadata such as domain, account surface, app,
+  process family, and path argument slots. Use
+  `unicli approvals list`, `revoke`, and `clear` to inspect or remove remembered
+  scopes. The file stores scope metadata; runtime args stay out of approval
+  memory.
+- Local deny rules live at `~/.unicli/permission-rules.json`, or at
+  `UNICLI_PERMISSION_RULES_PATH`. They match site, command, effect, capability
+  dimensions, and resource metadata, then block before `--yes` and remembered
+  approvals. Runtime guards also check fetched domains, browser navigation
+  targets, download and output paths, and subprocess executables before the
+  request, write, or process spawn happens.
 - Run recording is opt-in. Use `--record` or `UNICLI_RECORD_RUN=1` when you need
   append-only evidence under `~/.unicli/runs`.
 - CUA routes require a configured real backend. Declared-but-unavailable providers fail closed with structured errors.
 - User adapters and repairs live in `~/.unicli/adapters/`; committed adapters remain the package baseline.
-- If a site blocks automation or changes a private API, the correct behavior is a clear failure envelope, not a fabricated success.
+- If a site blocks automation or changes a private API, Uni-CLI returns a clear failure envelope.
 
 ## Development
 
@@ -314,5 +354,5 @@ npm run verify
 [Apache-2.0](./LICENSE)
 
 <p align="center">
-  <sub>v0.217.0 — Apollo · Lovell</sub>
+  <sub>v0.218.0 — Apollo · Cernan</sub>
 </p>

package/README.zh-CN.md

@@ -28,7 +28,7 @@
 </p>
 
 <p align="center">
-  <sub><!-- STATS:site_count -->235<!-- /STATS --> 个站点 · <!-- STATS:command_count -->1448<!-- /STATS --> 条命令 · <!-- STATS:pipeline_step_count -->59<!-- /STATS --> 个 pipeline step · <!-- STATS:test_count -->7471<!-- /STATS --> 个测试</sub>
+  <sub><!-- STATS:site_count -->237<!-- /STATS --> 个站点 · <!-- STATS:command_count -->1453<!-- /STATS --> 条命令 · <!-- STATS:pipeline_step_count -->101<!-- /STATS --> 个 pipeline step · <!-- STATS:test_count -->7842<!-- /STATS --> 个测试</sub>
 </p>
 
 <p align="center">
@@ -131,17 +131,17 @@ Prefer native CLI / JSON stream / MCP for agent runtimes. Use ACP as an editor c
 
 Uni-CLI 位于 Agent 应用之下，把软件表面收敛成 Agent 能发现、能执行、能记录、能修的命令。
 
-| 表面         | 能力                                                                            |
-| ------------ | ------------------------------------------------------------------------------- |
-| 网站和 API   | public、cookie、header、browser-intercept 等 adapter                            |
-| 浏览器自动化 | CDP 的 navigate、click、type、intercept、snapshot、extract、wait 等步骤         |
-| 桌面和 macOS | 系统命令、App adapter、截图、剪贴板、日历、亮度、本地工具                       |
-| 外部 CLI     | 58 个已登记的 passthrough bridge，支持安装和状态发现                            |
-| Agent 后端   | native CLI、JSON stream、MCP、ACP、HTTP API、OpenAI-compatible、bridge 路由矩阵 |
-| 操作策略     | `open`、`confirm`、`locked` profile，暴露 effect、risk、approval、capability    |
-| 执行证据     | 可选 run trace，浏览器动作的前后证据、移动检测、stale-ref 细节                  |
-| 输出         | v2 `AgentEnvelope`，支持 Markdown、JSON、YAML、CSV、compact                     |
-| 修复         | 错误里带 `adapter_path`、失败 `step`、是否可重试、修复建议和替代命令            |
+| 表面         | 能力                                                                                                                                                                       |
+| ------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| 网站和 API   | public、cookie、header、browser-intercept 等 adapter                                                                                                                       |
+| 浏览器自动化 | CDP 的 navigate、click、type、intercept、snapshot、extract、wait 等步骤                                                                                                    |
+| 桌面和 macOS | 系统命令、App adapter、实时 Shortcuts/App Intent 发现、截图、剪贴板、日历、亮度、本地工具                                                                                  |
+| 外部 CLI     | 58 个已登记的 passthrough bridge，支持安装和状态发现                                                                                                                       |
+| Agent 后端   | native CLI、JSON stream、MCP、ACP、HTTP API、OpenAI-compatible、bridge 路由矩阵                                                                                            |
+| 操作策略     | `open`、`confirm`、`locked` profile，暴露 effect/risk scope、本地 deny 规则、`--yes` 和持久审批记忆                                                                        |
+| 执行证据     | run trace 会记录环境快照，也能 probe/replay/compare 打分并输出结构化 gate 结果；浏览器 session lease 带 tab/auth 姿态，还支持 render-aware 证据、移动检测和 stale-ref 细节 |
+| 输出         | v2 `AgentEnvelope`，支持 Markdown、JSON、YAML、CSV、compact                                                                                                                |
+| 修复         | 错误里带 `adapter_path`、失败 `step`、是否可重试、修复建议和替代命令                                                                                                       |
 
 ## 给 Agent 的入口
 
@@ -150,6 +150,8 @@ Uni-CLI 位于 Agent 应用之下，把软件表面收敛成 Agent 能发现、
 ```bash
 unicli search "推特热门" --limit 5
 unicli twitter search "coding agents" -f json
+unicli macos app-actions --app WhatsApp -f json
+unicli macos automation-smoke -f json
 unicli repair twitter search
 ```
 
@@ -159,6 +161,16 @@ unicli repair twitter search
 UNICLI_OUTPUT=json unicli reddit hot --limit 10
 unicli hackernews top --limit 5 -f yaml
 unicli --record --permission-profile confirm twitter search "coding agents" -f json
+unicli runs list -f json
+unicli runs show <run_id> -f json
+unicli runs probe <run_id> -f json
+unicli runs replay <run_id> --permission-profile confirm --yes --min-score 1 --min-context-score 1 --min-overall-score 1 -f json
+unicli runs compare <run_id> <replay_run_id> -f json
+unicli runs compare <run_id> <replay_run_id> --min-score 1 --min-context-score 1 --min-overall-score 1 -f json
+unicli --permission-profile locked --yes --remember-approval word set-font "Inter"
+unicli approvals list -f json
+unicli approvals revoke <approval_key> -f json
+unicli browser evidence --render-aware --expect-domain example.com -f json
 ```
 
 协议入口：
@@ -175,7 +187,7 @@ ACP 作为编辑器和桥接兼容层保留。真正跑任务时，优先 native
 
 ## 覆盖范围
 
-数量不是重点，重点是每条命令都能搜索、可声明、可验证、可修。
+每条命令都能搜索、可声明、可验证、可修。
 
 | 领域          | 例子                                                                                                        |
 | ------------- | ----------------------------------------------------------------------------------------------------------- |
@@ -293,11 +305,20 @@ columns: [title, url]
 - Browser adapter 需要可连接的 Chrome/CDP。
 - Permission profile 是用户选择的运行时策略。默认是 `open`；更严格的
   `confirm` 和 `locked` profile 会要求 `--yes` 或 `UNICLI_APPROVE=1`。
+  `--yes` 加 `--remember-approval` 会把同一条命令的 capability 和资源 scope
+  记到 `~/.unicli/approvals.jsonl`。资源 scope 来自稳定 metadata，比如域名、
+  账号面、应用、进程族和路径参数槽。用 `unicli approvals list`、`revoke`、
+  `clear` 查看或移除已记住的 scope。文件只存 scope metadata，原始运行参数留在审批记忆之外。
+- 本地 deny 规则放在 `~/.unicli/permission-rules.json`，也可以用
+  `UNICLI_PERMISSION_RULES_PATH` 指定。规则按站点、命令、effect、capability
+  维度和资源 metadata 匹配，优先级高于 `--yes` 和已记住的审批。运行时还会检查
+  fetch 域名、浏览器跳转目标、下载和输出路径、子进程可执行文件，命中后在请求、写入或
+  启动进程之前停下。
 - Run recording 是显式启用能力。需要可审查证据时使用 `--record` 或
   `UNICLI_RECORD_RUN=1`，追加写入 `~/.unicli/runs`。
 - CUA 路由必须配置真实 backend。声明了但不可用的 provider 会失败关闭，并返回结构化错误。
 - 用户 adapter 和修复放在 `~/.unicli/adapters/`；包内 adapter 是基线。
-- 如果网站阻止自动化或私有 API 变了，正确行为是清楚失败，不是伪装成功。
+- 如果网站阻止自动化或私有 API 变了，Uni-CLI 会返回清楚的失败 envelope。
 
 ## 开发
 
@@ -313,5 +334,5 @@ npm run verify
 [Apache-2.0](./LICENSE)
 
 <p align="center">
-  <sub>v0.217.0 — Apollo · Lovell</sub>
+  <sub>v0.218.0 — Apollo · Cernan</sub>
 </p>

package/crates/unicli-atspi/Cargo.toml

@@ -0,0 +1,47 @@
+[package]
+name = "unicli-atspi"
+version.workspace = true
+edition.workspace = true
+rust-version.workspace = true
+license.workspace = true
+homepage.workspace = true
+repository.workspace = true
+
+[[bin]]
+name = "unicli-atspi"
+path = "src/main.rs"
+
+[dependencies]
+anyhow.workspace = true
+clap.workspace = true
+serde.workspace = true
+serde_json.workspace = true
+tokio.workspace = true
+tracing.workspace = true
+tracing-subscriber.workspace = true
+unicli-shared.workspace = true
+
+[target.'cfg(target_os = "linux")'.dependencies]
+atspi = { version = "0.29.0", default-features = false, features = [
+  "connection",
+  "proxies",
+  "wrappers",
+] }
+async-lock = "=3.4.1"
+async-signal = "=0.2.13"
+futures-lite = "=2.6.1"
+indexmap = "=2.11.3"
+proc-macro-crate = "=3.4.0"
+tempfile = "=3.22.0"
+toml_datetime = "=0.7.1"
+toml_edit = "=0.23.5"
+toml_parser = "=1.0.2"
+toml_writer = "=1.0.2"
+wasip2 = "=1.0.1"
+winnow = "=0.7.13"
+wit-bindgen = "=0.46.0"
+zbus = { version = "=5.11.0", features = ["async-io"], default-features = false }
+zbus_macros = "=5.11.0"
+zvariant = "=5.7.0"
+zvariant_derive = "=5.7.0"
+zvariant_utils = "=3.2.1"

package/crates/unicli-atspi/README.md

@@ -0,0 +1,6 @@
+# unicli-atspi
+
+Linux AT-SPI sidecar for Uni-CLI compute.
+
+P0 only establishes the buildable crate. The JSON-Lines stdio protocol and
+AT-SPI actions land in P4.