ax-grep-cli
Use ax-grep as a CLI-first page search and page-check tool before opening a browser. Best for agents that need compact semantic trees, source links, hidden app signals, forms, tables, citations, and browser handoff guidance.
معلومات المصدر
- المستودع
- hmmhmmhm/ax-grep
- آخر نشاط في المصدر
- ٢٠ يونيو ٢٠٢٦ في ١٠:٣٤
- لغة SKILL.md المكتشفة
- الإنجليزية
- النجوم
- ١٣
- التفرعات
- ٢
خيارات التثبيت
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
مراجعة ملفات المصدر
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
عرض SKILL.md
SKILL.md
تعليمات المصدر · معاينة للقراءة فقط- name
- ax-grep-cli
- description
- Use ax-grep as a CLI-first page search and page-check tool before opening a browser. Best for agents that need compact semantic trees, source links, hidden app signals, forms, tables, citations, and browser handoff guidance.
# ax-grep CLI
Use `ax-grep` before browser automation when a subagent needs to inspect a
search result, URL, HTML file, or captured page.
## First Command
```sh
ax-grep "https://example.com" --agent
```
If `ax-grep` is not on `PATH`, use the npm executable:
```sh
npx --yes ax-grep@latest "https://example.com" --agent
```
For smaller context:
```sh
ax-grep "https://example.com" --agent-brief
```
## Search
```sh
ax-grep --search "official docs pricing" --agent
```
Use the returned `agent.*CommandArgs` fields when continuing. They are JSON argv
arrays in execution order. Use `agent.*Command` when you need a shell string.
## Agent Policy
- Prefer `--agent` or `--agent-brief` over raw text output for subagents.
- Read `agent.executor`, `agent.handoff`, `agent.next`, `agent.readTargets`,
and any returned `agent.sourceChoices` or `agent.resultChoices` before
deciding.
- If `agent.executor.decision` is `return`, answer from the referenced
`readFrom` payload instead of opening a browser.
- If a URL follow-up is needed, run the provided `commandArgs` one command at a
time.
- Escalate to browser automation only when `needsBrowserHtml`,
`needsBrowserInteraction`, or the executor/handoff fields say it is needed.
- Do not run multiple browser-backed checks in parallel.
## Useful Patterns
Inspect a result and open the best match in one step:
```sh
ax-grep --search "site:example.com api reference" --open-result best --agent
```
If an engine rejects `best`, open the top ranked result directly:
```sh
ax-grep --search "site:example.com api reference" --open-result 1 --agent
```
Check a page for specific terms:
```sh
ax-grep "https://example.com/docs" --find "rate limit" --find "authentication" --agent
```
Read static HTML from another tool:
```sh
curl -fsSL "https://example.com" | ax-grep --stdin --agent
```
## What ax-grep Sees
`ax-grep --agent` can expose:
- semantic headings, landmarks, links, buttons, forms, tables, and lists
- source-like links and ranked search choices
- JSON-LD schema facts, metadata, policies, topics, provenance, and citations
- hydration URLs, API endpoints, runtime hints, app config, and mobile hints
- recommended next commands and browser handoff reasons
Treat the full JSON payload as source of truth; top-level shortcuts are for
fast routing.
عرض على GitHub