grok-search
Skill0 starsChanged 3 months agoMITView on GitHub
Real-time web + X (Twitter) search powered by xAI's Grok agentic search tools. Grok does the searching server-side and returns ranked results with citations. This is a deliberate, on-demand capability — it does not change your default web search backend. Use it when Grok's real-time reach or native access to X/Twitter is what the moment calls for; keep using the normal web_search tool for everything else. Load this skill when: - The user says "use Grok to search", "ask Grok",…
- Reads credentials
11 other projects use a version of this.
Adopted by
Most recently taken up by
- dengyie/awesome-skillstheir own version, first seen Oct 31, 2026
- BlueOcean223/grok-searchtheir own version, first seen Oct 27, 2026
- tanuuuuuuu/dotfilestheir own version, first seen Oct 20, 2026
- Frankieli123/grok-skilltheir own version, first seen Oct 8, 2026
- kaichen/agent-skillsettheir own version, first seen Sep 15, 2026
- Dianel555/DSkillstheir own version, first seen Sep 6, 2026
Other versions11 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 11
- Frankieli123/grok-skill★ 182same name
- jikssha/openclaw-dae-skills★ 98same name
- dengyie/awesome-skills★ 78same name
- Dianel555/DSkills★ 65same name
- BlueOcean223/grok-search★ 59same name
- Autsunset/grok-search★ 19same name
- JinchengGao-Infty/agent-loadout★ 3same name
- kaichen/agent-skillset★ 1same name
- yutakobayashidev/skills★ 1same name
- Lanfei/grok-search-skill★ 0same name
1 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository36 other files it gives its agents.
- check-upstream-firstSkill
- data-verificationSkill
- deep-diveSkill
- delegation-handoffSkill
- is-it-really-brokenSkill
- keep-goingSkill
- learn-from-conversationsSkill
- memory-cleanupSkill
- moa-solveSkill
- mob-checkSkill
- multi-reviewSkill
- project-stewardSkill
- prove-the-setting-worksSkill
- recallSkill
- reportSkill
- robustify-doctorSkill
- skill-librarianSkill
- trust-frameworkSkill
- address-pr-commentsSkill
- diagram-renderingSkill
- find-the-real-bottleneckSkill
- live-database-maintenanceSkill
- mini-appSkill
- pr-review-sweepSkill
- scheduled-job-runnerSkill
- stop-the-noiseSkill
- upgrade-a-live-service-safelySkill
- vendor-switch-checkSkill
- writing-for-non-engineersSkill
- email-stewardSkill
- fieldySkill
- google-docsSkill
- google-sheetsSkill
- google-slidesSkill
- imessage-bluebubblesSkill
- vapi-callsSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 2 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 9 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 39
Do not reach for this for ordinary web research — the default web_search provider is better for general-purpose, index-backed queries and …
- Distinctive: line 50
If the key is missing, the script exits with a clear JSON error.
- Very specific: line 94
If results is empty but citations has URLs, output the citations as a simple bulleted list and tell the user Grok found sources but didn't …
- Very specific: line 97
If the script errors (non-zero exit, JSON error on stderr), report the error plainly.
- Distinctive: line 108
Treat the descriptions as Grok's summaries, not verbatim source text.
- Very specific: line 129
Don't wire it as
web.backend: If you want Grok as a profile's default search, that's the built-in xai backend instead — a different, …
“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Grok Search
Real-time web + X (Twitter) search powered by xAI's Grok agentic search tools. Grok does the searching server-side and returns ranked results with citations.
This is a deliberate, on-demand capability — it does not change your default web
search backend. Use it when Grok's real-time reach or native access to X/Twitter is what
the moment calls for; keep using the normal web_search tool for everything else.
When to load
Load this skill when:
- The user says "use Grok to search", "ask Grok", "search X / Twitter", or names Grok/X explicitly for a lookup.
- The request is real-time / breaking ("what's happening with X right now", "latest reactions to Y") where a general index lags.
- The request is X/Twitter-native ("what are people posting about Z", "what's the sentiment on X about W", "find the thread where…").
Do not reach for this for ordinary web research — the default web_search provider
is better for general-purpose, index-backed queries and does not carry Grok's per-call
cost.
Prerequisites
XAI_API_KEYset in the profile's~/.hermes/.env(an xAI developer key from the xAI console). The script also accepts--api-key, but prefer the env var — a key on the command line lands in shell history and process listings.- Python 3 (stdlib only — no packages to install).
If the key is missing, the script exits with a clear JSON error; tell the user a key is needed and stop, don't fall back silently.
Usage
The helper is scripts/grok_search.py with two subcommands. It prints a JSON object to
stdout: {query, mode, results: [{title, url, description, source}], citations: [...]}.
Always shell-quote the query — wrap it in single quotes so punctuation, double
quotes, and shell metacharacters in the user's phrasing can't break the command or be
interpreted by the shell. If the query itself contains a single quote, escape it
('\'').
Web (real-time web search)
python3 scripts/grok_search.py web '<query>' --limit 5Optional:
--recency-days D— restrict to roughly the last D days (translated to afrom_dateof today − D, which is xAI's supported recency mechanism)--allowed-domains a.com,b.com— only these (max 5)--excluded-domains a.com,b.com— never these (max 5; mutually exclusive with allowed)--model M— override the Grok model (default is a reasoning model required for agentic tools)
X / Twitter search
python3 scripts/grok_search.py x '<query>' --limit 5Optional:
--handles elonmusk,xai— only posts from these handles (max 20)--from-date YYYY-MM-DD/--to-date YYYY-MM-DD— date window
How to present results
- Summarize the top results in your own words; cite the URLs. For X results the URLs are
x.com/<handle>/status/...— attribute to the handle when useful. - If
resultsis empty butcitationshas URLs, output the citations as a simple bulleted list and tell the user Grok found sources but didn't return structured summaries — don't invent descriptions for them. - If the script errors (non-zero exit, JSON
erroron stderr), report the error plainly. Do not invent results.
Trust & safety
Grok is an LLM, not a search index — it generates the result URLs it returns. That means:
- A query built from untrusted input can in principle steer Grok toward attacker-chosen URLs. Validate any URL before fetching it downstream, exactly as you would treat any model-generated link.
- Treat the descriptions as Grok's summaries, not verbatim source text. When accuracy matters, fetch the cited URL and confirm.
Cost
xAI bills Grok's built-in search tools per call (a flat per-1k-calls fee on top of token usage). This is why the skill is on-demand, not the default backend — reach for it when its real-time / X reach earns the call, not for routine lookups.
Pitfalls
- Reasoning model required. The agentic
web_search/x_searchtools need a reasoning-capable Grok model. The default (grok-4.5, the current flagship) supports them. If a non-reasoning model is passed via--model, the call may return no tool results. Stick with the default unless you know the target model supports agentic tools. - Not an extractor. This returns search results (title/url/description), not full
page content. To read a normal web page, pass the URL to the built-in
web_extracttool. Exception:x.comURLs —web_extractalmost always fails on them (auth-walls, JS rendering), so rely on Grok's own descriptions for X posts rather than trying to extract them. - Don't wire it as
web.backend. If you want Grok as a profile's default search, that's the built-inxaibackend instead — a different, deliberate choice with the same trust caveats. This skill deliberately leaves the default untouched.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

