⏳ This skill is pending AI review.
Scores will appear once the review pipeline completes.
giasip-research
Use when the user asks to research, investigate, compare competitors, study a market or industry, check the current status of a regulation or policy, search literature, or simply look something up that needs external sources. Any question that requires fetching outside information triggers this skill. To verify specific claims in an existing report, use a dedicated verification pass instead.
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
// README
giasip-skills
Turn a research question into a report with source links and an explicit list of what still needs checking. GiaSip Research is a reusable workflow for Codex and Claude Code. It splits the question across complementary research facets, then brings the findings together in one report.md.
简体中文 · Try Research · Example request and output
What you get
- An answer you can inspect: findings linked to their sources, with unresolved questions in a separate To verify section.
- A repeatable research workflow: complementary facets, bounded searching, and one report instead of disconnected search results.
- A clear boundary: a source link is not proof that a claim is true. Verification remains a separate step.
The repository also includes GiaSip Dispatch, multi-model dispatch tooling for Claude Code. The Codex plugin contains Research only.
35-second walkthrough

Chinese captions with English report excerpts. This edited walkthrough shows the output of one documentation-based research task comparing VHS with asciinema + agg. It is not a live screen recording or a speed benchmark. The task did not install or test those tools; its report keeps those checks explicit. Try the same question.
| Skill | What it gives you |
|---|---|
| giasip-research | Splits the question into 2–3 complementary facets, runs one sub-agent per facet in parallel (each capped at 15 searches/fetches, gathering only, never concluding), then writes a single report.md — answers first, every entity found with its URL, body under 200 lines, closing with a "To verify" section. Every fact carries a source URL; "not found" beats a guess from memory; a verification pass runs before anything ships under a real name, a price, or a legal/financial conclusion. |
| giasip-dispatch | A multi-model dispatcher — sends a task or prompt to other AI models (Codex / Gemini / Kimi / DeepSeek / Doubao / Qwen / GLM / MiniMax) and retrieves results. Start with one aggregator key (OpenRouter overseas / SiliconFlow in China) instead of signing up per vendor. Includes complexity-routing guidelines (API vs CLI vs SubAgent, single vs multi), but the final model choice is left to your agent's judgment. |
Quick Start
-
Install Research for your host:
# Claude Code npx skills add GiaSip/giasip-skills --global --skill giasip-research --agent claude-code --yes # Codex npx skills add GiaSip/giasip-skills --global --skill giasip-research --agent codex --yes # Or install the namespaced Codex Plugin codex plugin marketplace add GiaSip/giasip-skills codex plugin add giasip@giasip-skills -
Try it:
- Claude Code:
/giasip-research Research the current state of humanoid robot regulations - Codex:
$giasip-research Research the current state of humanoid robot regulations - Codex Plugin:
$giasip:research Research the current state of humanoid robot regulations - Or simply describe the research task in either host.
- Claude Code:
Installation
Choose one Codex distribution mode. Installing both is supported, but normally unnecessary.
Option 1: Standalone skill with npx skills add
# Claude Code: install all skills globally
npx skills add GiaSip/giasip-skills --global --skill '*' --agent claude-code --yes
# Claude Code: install GiaSip Research only
npx skills add GiaSip/giasip-skills --global --skill giasip-research --agent claude-code --yes
# Codex: install GiaSip Research only
npx skills add GiaSip/giasip-skills --global --skill giasip-research --agent codex --yes
# List available skills in this repo
npx skills add GiaSip/giasip-skills -l
Standalone invocation: /giasip-research in Claude Code or $giasip-research in Codex.
Option 2: Codex Plugin with the GiaSip namespace
codex plugin marketplace add GiaSip/giasip-skills
codex plugin add giasip@giasip-skills
Plugin invocation: $giasip:research.
The Codex Plugin intentionally bundles Research only. giasip-dispatch is not bundled because it remains Claude Code-native. See Codex Plugin architecture and maintenance.
Option 3: Claude Code plugin (Claude Code only)
/plugin marketplace add GiaSip/giasip-skills
/plugin install giasip-skills@giasip-skills
Option 4: git clone and copy
git clone https://github.com/GiaSip/giasip-skills
# Claude Code
cp -R giasip-skills/skills/giasip-research ~/.claude/skills/giasip-research
cp -R giasip-skills/skills/giasip-dispatch ~/.claude/skills/giasip-dispatch
# Codex / Agent Skills-compatible hosts
cp -R giasip-skills/skills/giasip-research ~/.agents/skills/giasip-research
Both standalone and plugin installs use the same Research behavior — only the invocation surface changes:
- Standalone (Options 1 & 4):
/giasip-researchin Claude Code,$giasip-researchin Codex;giasip-dispatchis/giasip-dispatchin Claude Code.- Codex Plugin (Option 2):
$giasip:research.- Claude Code plugin (Option 3): skills are namespaced by the plugin —
/giasip-skills:giasip-researchand/giasip-skills:giasip-dispatch.
Distribution Structure
skills/giasip-research/ # Standalone Claude/Codex install target
├── SKILL.md # behavioral source of truth
└── FEEDBACK.md # point-of-use human feedback log
plugins/giasip/ # Codex Plugin package
├── .codex-plugin/plugin.json
└── skills/research/SKILL.md # manually kept copy, name: research
.agents/plugins/marketplace.json # Repo marketplace for Codex
skills/giasip-dispatch/ # Claude Code-native dispatcher
giasip-research — Dependencies
Zero external dependencies — works out of the box. It uses your host's native web search / fetch tools and sub-agents (Claude Code's WebSearch / WebFetch / SubAgents, or Codex's equivalent). No configuration file to fill in.
giasip-dispatch — Dependencies
Pick the path that fits. Most users want the easy path — one aggregator key, no per-vendor signup.
1. Easy path — one aggregator key (recommended)
One key routes to many models through an OpenAI-compatible aggregator. Choose by region, drop one .env in ~/.config/ai-keys/, set the provider once, and every alias your chosen provider supports just works.
| Region | Provider | File | Content | Covers |
|---|---|---|---|---|
| Overseas | OpenRouter | openrouter.env | export OPENROUTER_API_KEY=... | DeepSeek / Qwen / GLM / Kimi / MiniMax + Claude / GPT / Gemini |
| China | SiliconFlow 硅基流动 | siliconflow.env | export SILICONFLOW_API_KEY=... | DeepSeek / Qwen / GLM / Kimi / MiniMax |
export DISPATCH_PROVIDER=openrouter # or: siliconflow
~/.claude/skills/giasip-dispatch/scripts/api-dispatch.sh --model deepseek "Hello"
- Get a key: OpenRouter → https://openrouter.ai/keys · SiliconFlow → https://siliconflow.cn
- Provider resolution:
--via <provider>flag >$DISPATCH_PROVIDERenv >direct. Escape hatch for any model the alias table misses:--model-id <raw>(e.g.--via openrouter --model-id anthropic/claude-3.7-sonnet). - Caveats: OpenRouter passes model inference pricing through at parity (no per-token markup) but takes ~5.5% on credit top-ups, and needs a VPN in mainland China. SiliconFlow is China-direct but open-source/domestic o
// HOW IT'S BUILT
KEY FILES