chromeapi
Capture a site's internal API routes from your REAL logged-in Chrome and replay them browserless (kuri/unbrowse-style, all local). The deeper-level API layer on top of fcdp. Invoke as /chromeapi when the user wants to record the XHR/fetch calls a page makes and re-fire them as plain HTTP with their real cookies — no browser, no telemetry. /chrome (fcdp) remains the tool for ordinary page driving.
- License 许可
- MIT
- 最近更新
- 2026-11-06
- Stars
- ★ 4
- Forks
- 0
01 前言 Front Matter
- name
- chromeapi
- description
- Capture a site's internal API routes from your REAL logged-in Chrome and replay them browserless (kuri/unbrowse-style, all local). The deeper-level API layer on top of fcdp. Invoke as /chromeapi when the user wants to record the XHR/fetch calls a page makes and re-fire them as plain HTTP with their real cookies — no browser, no telemetry. /chrome (fcdp) remains the tool for ordinary page driving.
- user-invocable
- true
03 SKILL.md 正文
/chromeapi — capture & replay a site's internal API (on your real Chrome)
The deeper-level API layer built on top of /chrome (fcdp). Same real logged-in
Default-profile Chrome — but instead of clicking/reading a page, it records the internal
API calls the page fires into a manifest, then replays them browserless as plain curl
using your real Chrome cookies. This is the local, no-telemetry equivalent of unbrowse/kuri's
capture→replay trick (reverse-engineered 2026-07-22; see memory reference_fcdp_full_cdp.md).
/chrome (fcdp) stays exactly as-is for ordinary driving (open/read/click/type/js/pdf/…).
/chromeapi is when the goal is "find the API call and replay it without a browser."
Step 0 — does the site PUBLISH its API? Check before you capture anything
Run this first, every time. A capture window is a guess: you get only the endpoints the page happened to fire in those N seconds, method+URL only, no schema. If the site publishes a spec, that's the entire API surface with parameter names, types and response schemas — strictly better data, in one request, with no browser.
Note this is the API analog of /chrome's sitemap-first rule, not sitemap.xml itself.
A sitemap enumerates pages; it tells you nothing about an API. The right artifacts here
are OpenAPI/Swagger/GraphQL. (Do still glance at robots.txt — it sometimes names an
/api-docs path.)
# Prefer the a logged-in browser CLI probe (robots/sitemap/OpenAPI/headers/scripts):
a logged-in browser CLI inspect https://<host>
# Old unbrowse MCP spec-discover is UNINSTALLED — do not invoke it.
# Unpublished / cookie-gated: a logged-in browser CLI capture <url> [--fcdp]
# or ~/tools/fcdp-api/fapi capture <url> [secs]
# HAR distill: a logged-in browser CLI distill <file.har>
Decision rule:
- Spec found → read it. You now have every route, its params and its response shape.
Skip
fapi captureentirely and go straight tocurl(still with real cookies for authenticated routes). This is also the only way to learn about endpoints the UI never calls. - GraphQL (
/graphqlresponds) → introspection may be enabled; that's the whole schema. Introspection is often disabled in production — if so, fall through to capture. - Nothing published (the common case for consumer web apps) → proceed to
fapi capturebelow. That's the normal path, not a failure.
A 404 on all of the above is a 5-second answer, so there's no reason to skip it — and when
it hits, it saves the capture cycle and gives you schemas fapi structurally cannot infer
(see the "Gap vs unbrowse/kuri" note below: fcdp intercept surfaces method+URL, not bodies).
The tool: fapi
~/tools/fcdp-api/fapi (also ~/tools/fcdp/fapi). Built on fcdp intercept (capture) +
~/tools/cookies-txt (real-profile cookies) + curl (browserless replay). Nothing leaves the machine.
fapi capture <url> [secs] # open a bg tab in REAL Chrome, intercept the API calls it fires -> manifest
fapi list # list captured domains
fapi show <domain> # print the endpoint manifest (JSON)
fapi replay <domain> [n] # browserless curl replay of GET endpoint(s), with real Chrome cookies
Manifests live in ~/.fcdp-api/<domain>.json: {domain, trigger, captured_at, endpoints:[{method,url,idempotency}]}.
How to run it
- Capture — pick a URL that fires the API you want (the page, or the API URL itself). Give it
enough seconds to let the page's XHRs fire (default 8):
Report the endpoints it found (~/tools/fcdp-api/fapi capture "https://app.example.com/dashboard" 8fapi show <domain>). If it found 0, the page may fire its XHRs later than the window — re-run with more seconds, or pointcapturedirectly at the API URL. - Replay — re-fire the safe (GET) endpoints without a browser, using the user's real cookies:
This is step 3 of the web-lookup ladder (cached endpoint replay, no live browser) done against the user's OWN session.~/tools/fcdp-api/fapi replay app.example.com
Rules & safety
- GET-only auto-replay.
fapi replayfires onlyGETendpoints. POST/PUT/DELETE/PATCH are listed but NEVER auto-fired — a mutation is an outward, irreversible action: show the exact request and get explicit per-action approval, then fire it by hand withcurl. - Real cookies, local only.
~/tools/cookies-txtreads the Default profile's cookie SQLite + Keychain. Cookies never leave the machine. Only capture/replay sites the user is lawfully using their own account on. - One debugger per tab.
fapi captureusesfcdp intercept, which attacheschrome.debuggerto a fresh background tab and closes it after — it won't fight a tab you're driving with/chrome. - Gap vs unbrowse/kuri (by design): stores method+URL only — no response-schema inference or
ranked semantic resolve, because
fcdp interceptsurfaces method+URL, not response bodies. If schema inference is ever wanted, addNetwork.getResponseBodycapture to fcdp. Better than kuri in the way that matters here: it runs against the user's REAL logins (no separate auth step).
When to use which
| Want to… | Use |
|---|---|
| inspect / fetch / capture / GET-replay / drive as one stack | a logged-in browser CLI (preferred for new work) |
| open/read/click/type/screenshot/PDF a page, run JS | /chrome (fcdp) |
| record a site's internal API calls and replay them browserless | a logged-in browser CLI capture + replay, or this skill's fapi |
| decompile a binary / RE how some app talks to its API | /decompile |
| turn a captured HAR/JSON into a repeatable assert (jsonpath on unwrap, status, body keys) | hurl --test file.hurl (installed 2026-08-14). Capture with fhar/fapi/a logged-in browser CLI, flatten with gron to find the path, then write the Hurl file. Never commit the raw HAR. |
Full RE writeup of how this mirrors unbrowse/kuri: memory reference_fcdp_full_cdp.md.
04 同仓库更多 Skills
zi-search
暂无简介
visualise
Render inline interactive visuals — SVG diagrams, HTML widgets, charts, and explainers — directly in the conversation. Use this skill whenever the user asks to visualize, diagram, chart, illustrate, or explain something visually, or when an explanation would genuinely benefit from a spatial/interactive diagram rather than text. Also triggers for: flowcharts, architecture diagrams, data visualizations, interactive explainers, comparison layouts, UI mockups, and any request containing 'show me', 'draw', 'map out', 'visualize', or 'diagram'. Even when the user doesn't explicitly ask for a visual, use this skill proactively when the topic has spatial, sequential, or systemic relationships that a diagram would clarify better than prose.
video
When the user wants to create, generate, or produce video content using AI tools or programmatic frameworks. Also use when the user mentions 'video production,' 'AI video,' 'Remotion,' 'Hyperframes,' 'HeyGen,' 'Synthesia,' 'Veo,' 'Runway,' 'Kling,' 'Pika,' 'video generation,' 'AI avatar,' 'talking head video,' 'programmatic video,' 'video template,' 'explainer video,' 'product demo video,' 'video pipeline,' or 'make me a video.' Use this for video creation, generation, and production workflows. For video content strategy and what to post, see social-content. For paid video ad creative, see ad-creative.
manim-video
Production pipeline for mathematical and technical animations using Manim Community Edition. Creates 3Blue1Brown-style explainer videos, algorithm visualizations, equation derivations, architecture diagrams, and data stories. Use when users request: animated explanations, math animations, concept visualizations, algorithm walkthroughs, technical explainers, 3Blue1Brown style videos, or any programmatic animation with geometric/mathematical content.
video-use
Edit any video by conversation. Transcribe, cut, color grade, generate overlay animations, burn subtitles — for talking heads, montages, tutorials, travel, interviews. No presets, no menus. Ask questions, confirm the plan, execute, iterate, persist. Production-correctness rules are hard; everything else is artistic freedom.
ui-improver
暂无简介