research

12 / 100 · High risk

Only if you trust the author. Some of what it can do is expected: a research skill usually looks things up on the internet.

By lev-os/agents (GitHub) · research · 27 stars · scanned at commit 201a805

Use when current external facts, source-backed comparisons, investigations, or recent community evidence are needed.

Works with any agent that supports SKILL.md skills.

What this skill can do

Can call paid AI services billed to you

The author does not mention this.

Found in 23 places.

skills/research/apps/research-cli/src/backends/perplexity.rs, line 48.post("https://api.perplexity.ai/chat/completions")
skills/research/backends/grok-web-twitter-realtime/BACKEND.md, line 23curl -s https://api.x.ai/v1/responses \
skills/research/backends/grok-web-twitter-realtime/scripts/grok-search.sh, line 28curl -s https://api.x.ai/v1/chat/completions \

Show 17 more

skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 276- xAI Grok 4.1 API support (`--model grok-4.1` / alias `grok`): defaults to `https://api.x.ai/v1`, uses `XAI_API_KEY`, maps search to `web_search`, and includes
skills/research/backends/oracle-ceo-gpt5pro/references/releases.md, line 378- Add xAI Grok 4.1 API support (`--model grok-4.1` / alias `grok`), defaults to `https://api.x.ai/v1`, uses `XAI_API_KEY`, and maps search to `web_search` with
skills/research/references/last30days/METHOD.md, line 2065- Legacy: Sends search queries to OpenAI's Responses API (`api.openai.com`) for Reddit discovery (fallback if no SCRAPECREATORS_API_KEY)
skills/research/references/last30days/METHOD.md, line 2066- Sends search queries to X/Twitter via optional user-provided `AUTH_TOKEN`/`CT0` env vars, explicit browser-cookie opt-in (`FROM_BROWSER` or setup consent), xA
skills/research/references/last30days/METHOD.md, line 2071- Optionally sends search queries to Brave Search API, Parallel AI API, Perplexity API (`api.perplexity.ai`), or OpenRouter API for web search / synthesis
skills/research/references/last30days/scripts/evaluate_search_quality.py, line 48GEMINI_API_URL = "https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={api_key}"
skills/research/references/last30days/scripts/lib/perplexity.py, line 19OPENROUTER_URL = "https://openrouter.ai/api/v1/chat/completions"
skills/research/references/last30days/scripts/lib/perplexity.py, line 20PERPLEXITY_AGENT_URL = "https://api.perplexity.ai/v1/agent"
skills/research/references/last30days/scripts/lib/perplexity.py, line 21PERPLEXITY_SEARCH_URL = "https://api.perplexity.ai/search"
skills/research/references/last30days/scripts/lib/prescriptions.py, line 82"api.x.ai), or add XQUIK_API_KEY to your .env (get key at xquik.com)"
skills/research/references/last30days/scripts/lib/providers.py, line 18GEMINI_URL = "https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={api_key}"
skills/research/references/last30days/scripts/lib/providers.py, line 19OPENAI_RESPONSES_URL = "https://api.openai.com/v1/responses"
skills/research/references/last30days/scripts/lib/providers.py, line 20XAI_RESPONSES_URL = "https://api.x.ai/v1/responses"
skills/research/references/last30days/scripts/lib/providers.py, line 21OPENROUTER_URL = "https://openrouter.ai/api/v1/chat/completions"
skills/research/references/last30days/scripts/lib/transcribe.py, line 36"openai": "https://api.openai.com/v1/audio/transcriptions",
skills/research/references/last30days/scripts/lib/xai_x.py, line 28XAI_RESPONSES_URL = "https://api.x.ai/v1/responses"
skills/research/references/legacy-last30days-backend/scripts/lib/models.py, line 9OPENAI_MODELS_URL = "https://api.openai.com/v1/models"

and 3 more places

Downloads and installs other software

The author does not mention this.

Found in 95 places.

skills/research/apps/research-hub/INTEGRATION.md, line 152npm install -g @valyu/cli
skills/research/apps/research-hub/INTEGRATION.md, line 153npm install -g firecrawl-cli
skills/research/apps/research-hub/INTEGRATION.md, line 154brew install brave-cli # or equivalent

Show 17 more

skills/research/backends/firecrawl-scrape-extract/rules/install.md, line 12npm install -g firecrawl-cli
skills/research/backends/firecrawl-scrape-extract/rules/install.md, line 762. Try: `npx firecrawl-cli --version`
skills/research/backends/firecrawl-scrape-extract/rules/install.md, line 773. Or reinstall: `npm install -g firecrawl-cli`
skills/research/backends/firecrawl-scrape-extract/rules/install.md, line 85sudo npm install -g firecrawl-cli
skills/research/backends/oracle-ceo-gpt5pro/BACKEND.md, line 78npm install -g @steipete/oracle
skills/research/backends/oracle-ceo-gpt5pro/BACKEND.md, line 81brew install steipete/tap/oracle
skills/research/backends/oracle-ceo-gpt5pro/BACKEND.md, line 84npx -y @steipete/oracle …
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 187- Docs: README now explicitly warns against `pnpx @steipete/oracle` (pnpx cache breaks sqlite bindings); use `npx -y @steipete/oracle` instead. Thanks Xuanwo fo
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 271- Publish metadata now declares Node >=20 (engines/devEngines) and drops the implicit bun runtime so `npx @steipete/oracle` no longer fails with EBADDEVENGINES
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 288- MCP: `npx @steipete/oracle oracle-mcp` now routes directly to the MCP server (even when npx defaults to the CLI binary) and keeps stdout JSON-only for Cursor/
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 289- Added the missing `@anthropic-ai/tokenizer` runtime dependency so `npx @steipete/oracle oracle-mcp` starts cleanly.
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 383"command": "npx",
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 393"oracle": { "type": "stdio", "command": "npx", "args": ["-y", "@steipete/oracle", "oracle-mcp"] }
skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 469- Positional prompt shorthand: `oracle "prompt here"` (and `npx -y @steipete/oracle "..."`) now maps the positional argument to `--prompt` automatically.
skills/research/backends/oracle-ceo-gpt5pro/references/README.md, line 18Install globally: `npm install -g @steipete/oracle`
skills/research/backends/oracle-ceo-gpt5pro/references/README.md, line 19Homebrew: `brew install steipete/tap/oracle`
skills/research/backends/oracle-ceo-gpt5pro/references/README.md, line 21Requires Node 22+. Or use `npx -y @steipete/oracle …` (or pnpx).

and 75 more places

Changes files outside its own folder

The author does not mention this.

Found in 21 places.

skills/research/apps/research-hub/TECHNICAL_MINIMALIST_COMPLETE.md, line 146cp ~/.agents/skills/lev-research/apps/research-hub/src/components/Dashboard.tsx.backup ~/.agents/skills/lev-research/apps/research-hub/src/components/Dashboard.
skills/research/backends/firecrawl-scrape-extract/rules/install.md, line 88mkdir ~/.npm-global
skills/research/backends/oracle-ceo-gpt5pro/BACKEND.md, line 203mkdir -p ~/.codex/skills

Show 17 more

skills/research/backends/oracle-ceo-gpt5pro/BACKEND.md, line 204cp -R skills/oracle ~/.codex/skills/oracle
skills/research/backends/oracle-ceo-gpt5pro/references/README.md, line 70- `mkdir -p ~/.codex/skills`
skills/research/backends/oracle-ceo-gpt5pro/references/README.md, line 71- `cp -R skills/oracle ~/.codex/skills/oracle`
skills/research/backends/valyu-recursive-confidence/IMPLEMENTATION.md, line 88echo "VALYU_API_KEY=your-key" > ~/.valyu/.env
skills/research/backends/valyu-recursive-confidence/IMPLEMENTATION.md, line 2522. Configure API key: `echo "VALYU_API_KEY=key" > ~/.valyu/.env`
skills/research/backends/valyu-recursive-confidence/cli/README.md, line 13echo "VALYU_API_KEY=your-key" > ~/.valyu/.env
skills/research/backends/valyu-recursive-confidence/cli/README.md, line 396echo "VALYU_API_KEY=your-key" > ~/.valyu/.env
skills/research/backends/valyu-recursive-confidence/cli/install.sh, line 34mkdir -p ~/.valyu
skills/research/references/last30days/METHOD.md, line 438**If the user picks Skip for now:** write `SETUP_COMPLETE=true` to `~/.config/last30days/.env` (append-only; run `mkdir -p ~/.config/last30days && touch ~/.conf
skills/research/references/last30days/METHOD.md, line 5724. If it doesn't exist: `mkdir -p ~/.config/last30days && touch ~/.config/last30days/.env`
skills/research/references/last30days/scripts/lib/prescriptions.py, line 229"add the install directory (default ~/.local/bin) to the PATH the "
skills/research/references/last30days/scripts/lib/render.py, line 31Per-harness skill install dirs (`~/.claude/skills`, `~/.codex/skills`, `~/.agents/skills`,
skills/research/references/last30days/scripts/test_device_auth.py, line 122print(f"\nTo use it: echo 'SCRAPECREATORS_API_KEY={api_key}' >> ~/.config/last30days/.env")
skills/research/references/legacy-last30days-backend/BACKEND.md, line 66mkdir -p ~/.config/last30days
skills/research/references/legacy-last30days-backend/BACKEND.md, line 67cat > ~/.config/last30days/.env << 'ENVEOF'
skills/research/references/legacy-last30days-backend/README.md, line 16mkdir -p ~/.config/last30days
skills/research/references/legacy-last30days-backend/README.md, line 17cat > ~/.config/last30days/.env << 'EOF'

and 1 more places

Sends data to these websites: registry.npmjs.org, api.tavily.com, api.exa.ai

This is disclosed by the author in the skill's SKILL.md or README. Expected for a research skill.

Found in 210 places.

skills/research/backends/brave-web-quick/cli/package-lock.json, line 11"axios": "^1.6.5",
skills/research/backends/brave-web-quick/cli/package-lock.json, line 496"node_modules/axios": {
skills/research/backends/brave-web-quick/cli/package-lock.json, line 498"resolved": "https://registry.npmjs.org/axios/-/axios-1.13.4.tgz",

Show 17 more

skills/research/backends/brave-web-quick/cli/package.json, line 14"axios": "^1.6.5",
skills/research/backends/brave-web-quick/cli/src/brave.ts, line 4import axios, { AxiosInstance } from 'axios';
skills/research/backends/brave-web-quick/cli/src/brave.ts, line 26return axios.create({
skills/research/backends/brave-web-quick/content.js, line 37const response = await fetch(url, {
skills/research/backends/brave-web-quick/search.js, line 38const response = await fetch(url, {
skills/research/backends/brave-web-quick/search.js, line 112const response = await fetch(url, {
skills/research/backends/deep-multi-query-synthesis/scripts/research.mjs, line 89const resp = await fetch("https://api.tavily.com/search", {
skills/research/backends/exa-neural-linkedin-github/scripts/code.sh, line 31curl -s -X POST 'https://api.exa.ai/search' \
skills/research/backends/exa-neural-linkedin-github/scripts/content.sh, line 36curl -s -X POST 'https://api.exa.ai/contents' \
skills/research/backends/exa-neural-linkedin-github/scripts/search.sh, line 82curl -s -X POST 'https://api.exa.ai/search' \
skills/research/backends/firecrawl-scrape-extract/cli/package-lock.json, line 11"axios": "^1.6.5",
skills/research/backends/firecrawl-scrape-extract/cli/package-lock.json, line 496"node_modules/axios": {
skills/research/backends/firecrawl-scrape-extract/cli/package-lock.json, line 498"resolved": "https://registry.npmjs.org/axios/-/axios-1.13.4.tgz",
skills/research/backends/firecrawl-scrape-extract/cli/package.json, line 15"axios": "^1.6.5",
skills/research/backends/firecrawl-scrape-extract/cli/src/firecrawl.ts, line 4import axios, { AxiosInstance } from 'axios';
skills/research/backends/firecrawl-scrape-extract/cli/src/firecrawl.ts, line 26return axios.create({
skills/research/backends/firecrawl-scrape-extract/cli/src/firecrawl.ts, line 95const client = axios.create({

and 190 more places

Runs commands on your computer

The author does not mention this.

Found in 50 places.

skills/research/backends/firecrawl-scrape-extract/BACKEND.md, line 325cat urls.txt | xargs -P 10 -I {} sh -c 'firecrawl scrape "{}" -o ".firecrawl/$(echo {} | md5).md"'
skills/research/backends/valyu-recursive-confidence/BACKEND.md, line 580import { exec } from 'child_process';
skills/research/backends/valyu-recursive-confidence/IMPLEMENTATION.md, line 216import { exec } from 'child_process';

Show 17 more

skills/research/references/last30days/scripts/box_chrome_login.py, line 200subprocess.Popen(recipe["command"], env=launch_env)
skills/research/references/last30days/scripts/evaluate_search_quality.py, line 347result = subprocess.run(
skills/research/references/last30days/scripts/evaluate_search_quality.py, line 372subprocess.run(
skills/research/references/last30days/scripts/evaluate_search_quality.py, line 390subprocess.run(
skills/research/references/last30days/scripts/lib/agentcookie.py, line 163result = subprocess.run(
skills/research/references/last30days/scripts/lib/agentcookie.py, line 169except subprocess.TimeoutExpired:
skills/research/references/last30days/scripts/lib/backends.py, line 274``subprocess.run([name, "--version"])``, and the whole-doctor-path test
skills/research/references/last30days/scripts/lib/backends.py, line 275patches ``subprocess.run`` to raise.
skills/research/references/last30days/scripts/lib/chrome_cookies.py, line 71result = subprocess.run(
skills/research/references/last30days/scripts/lib/chrome_cookies.py, line 88except subprocess.TimeoutExpired:
skills/research/references/last30days/scripts/lib/chrome_cookies.py, line 127result = subprocess.run(
skills/research/references/last30days/scripts/lib/chrome_cookies.py, line 160except subprocess.TimeoutExpired:
skills/research/references/last30days/scripts/lib/corpus.py, line 161except (OSError, subprocess.SubprocessError) as exc:
skills/research/references/last30days/scripts/lib/corpus.py, line 291completed = subprocess.run(
skills/research/references/last30days/scripts/lib/env.py, line 396result = subprocess.run(
skills/research/references/last30days/scripts/lib/env.py, line 403except (subprocess.TimeoutExpired, OSError):
skills/research/references/last30days/scripts/lib/env.py, line 444result = subprocess.run(

and 30 more places

Reads your saved credentials, such as SSH keys or .env files

The author does not mention this.

Found in 10 places.

skills/research/backends/oracle-ceo-gpt5pro/BACKEND.md, line 374. For API mode: `source ~/.env.local && oracle --engine api -p "..."`
skills/research/references/last30days/scripts/lib/env.py, line 433source like Keychain, so an explicit .env or process-env value still wins.
skills/research/references/last30days/scripts/lib/grok_x.py, line 322SCRAPECREATORS_API_KEY to ``$HOME/.config/last30days/.env``, and ``~/.ssh``

Show 7 more

skills/research/references/last30days/scripts/lib/grok_x.py, line 323and ``~/.aws`` sit alongside it. An empty cwd is not a boundary for a
skills/research/references/last30days/scripts/lib/x_envelope.py, line 113"~/.aws",
skills/research/references/last30days/scripts/lib/x_envelope.py, line 114"~/.ssh",
skills/research/references/last30days/scripts/lib/x_envelope.py, line 115"~/.netrc",
skills/research/references/last30days/scripts/lib/youtube_yt.py, line 311SSH alias (configured in ~/.ssh/config). On macOS hosts with Homebrew,
skills/research/references/legacy-last30days-backend/BACKEND.md, line 67cat > ~/.config/last30days/.env << 'ENVEOF'
skills/research/references/legacy-last30days-backend/README.md, line 17cat > ~/.config/last30days/.env << 'EOF'

Reads your browser cookies or saved passwords

The author does not mention this.

Found in 12 places.

skills/research/backends/oracle-ceo-gpt5pro/references/CHANGELOG.md, line 326- Browser cookie sync on Windows now copies the profile DB into a named temp directory with the expected `Cookies` filename so `chrome-cookies-secure` can read
skills/research/backends/oracle-ceo-gpt5pro/references/releases.md, line 442- Browser cookie sync on Windows now copies the profile DB into a named temp directory with the expected `Cookies` filename so `chrome-cookies-secure` can read
skills/research/references/last30days/scripts/lib/chrome_cookies.py, line 65Calls `security find-generic-password` which may trigger a system dialog

Show 9 more

skills/research/references/last30days/scripts/lib/cookie_extract.py, line 153"""Find the first directory that contains cookies.sqlite."""
skills/research/references/last30days/scripts/lib/cookie_extract.py, line 156if child.is_dir() and (child / "cookies.sqlite").is_file():
skills/research/references/last30days/scripts/lib/cookie_extract.py, line 168Firefox locks cookies.sqlite while running, so we copy first.
skills/research/references/last30days/scripts/lib/cookie_extract.py, line 235result = _query_cookies_db(default_profile / "cookies.sqlite", domain, cookie_names)
skills/research/references/last30days/scripts/lib/cookie_extract.py, line 246db = child / "cookies.sqlite"
skills/research/references/last30days/scripts/lib/cookie_extract.py, line 263Finds the default Firefox profile, copies cookies.sqlite to a temp file
skills/research/references/last30days/scripts/lib/safari_cookies.py, line 4Parses ~/Library/Cookies/Cookies.binarycookies (unencrypted binary format)
skills/research/references/last30days/scripts/setup-keychain.sh, line 77if security find-generic-password -a "$USER" -s "${PREFIX}${key}" >/dev/null 2>&1; then
skills/research/references/last30days/scripts/setup-keychain.sh, line 104if security find-generic-password -a "$USER" -s "${PREFIX}${key}" >/dev/null 2>&1; then

Safer alternatives

Low-risk skills in the same category (research) that look related. Check that one does what you need.

Install

Paste this into your terminal. It downloads exactly the version we scanned, not whatever is newest.

if [ -e ~/.agents/skills/'research' ] || [ -L ~/.agents/skills/'research' ]; then echo 'A skill folder named research already exists — not replacing it' >&2; false; else d=$(mktemp -d) && git clone --quiet --filter=blob:none --no-checkout 'https://github.com/lev-os/agents' "$d" && git -C "$d" checkout --quiet '201a80515e1dc58addfda3cd9978907de2c944a6' -- 'skills/research' && mkdir -p ~/.agents/skills && cp -R "$d"/'skills/research' ~/.agents/skills/'research' && rm -rf "$d"; fi

Codex reads ~/.agents/skills directly. For Claude Code, Cursor, Copilot, Gemini CLI, Antigravity or OpenCode, also link it. For Claude Code: ln -s ~/.agents/skills/'research' ~/.claude/skills/'research' (other agents: ~/.cursor/skills, ~/.copilot/skills, ~/.gemini/skills, ~/.gemini/config/skills, ~/.config/opencode/skills). Or install through the finder skill, which does both.

Badge for your README

The badge for lev-os/agents shows the score of its lowest-scoring skill.

SkillTrust badge for lev-os/agents

[![SkillTrust](https://zalavijaysinh.github.io/skilltrust/badges/lev-os/agents.svg)](https://zalavijaysinh.github.io/skilltrust/repos/lev-os/agents/)

Score history

Date	Version	Score
2026-11-07	`201a805`	12