perplexity-researchlisted
Install: claude install-skill rainearcher/quiver
# Perplexity Deep Research
Call the Perplexity Agent API (`POST /v1/agent`) to perform multi-step browsing research with citations. **Before** calling the API, this skill requires a structured intake + plan-approval loop so the agent uses the right sources, has the right context, and returns a report in a skimmable format.
## Methodology
This skill is a programmatic enforcement of the Growth Unhinged deep-research playbook for GTM: https://www.growthunhinged.com/p/deep-research-for-gtm. The article's five recommendations map 1:1 to the steps below:
| Article recommendation | Enforced by |
|---|---|
| Provide comprehensive context (goal, GTM/product, constraints, preferred sources) | Step 1 — five-dimension intake |
| Request a research plan first, approve before running | Step 2 — plan-approval gate (no API spend without approval) |
| Point to high-quality sources; prefer primary over secondary; cite everything | Format block — Sources table, primary-source priority, inline citations required |
| Specify report format; summaries and tables over prose (Pyramid Principle) | Format block — TL;DR first, comparison tables, conclusions before background |
| Structure the prompt (`<goal>`, `<context>`, `<sources>`, `<instructions>`) | `--instructions` block layout |
The per-call output-budget check (Step 2.5) and multi-call sequencing (Step 2.6) are additions on top of the article, to prevent the silent-truncation failure mode that large GTM plans (hook libraries, 30-day calend