deep-research

/deep-research - Deep Research

User invoked the deep-research skill with the following arguments:

$ARGUMENTS

Your task

Follow ${CLAUDE_SKILL_DIR}/SKILL.md. You are the researcher: scout the question, induce the outline from what retrieval returns, retrieve per section, keep every source in an evidence ledger, and write a cited report where each claim points back to a retrieved source. No claim reaches the report without a ledger source.

Use this command for tasks that need a sourced report — literature review, research landscape, entity investigation, trend comparison, or industry / market survey — not for a single lookup or a bare bibliography.

1. Parse $ARGUMENTS

• topic: required research topic. Preserve the user's wording. If absent or too vague, ask for a concrete topic (at most two questions, and only when different answers would change the scope).
• genre: optional, academic (default — literature reviews, landscapes, investigations) or industry (industry / market surveys: "行业调研", "市场格局", "竞争格局").

2. Pre-flight

Check the key without printing it:

[ -z "${AMINER_API_KEY:-}" ] && echo "AMINER_API_KEY missing" || echo "AMINER_API_KEY exists"

If missing, stop and ask the user to set AMINER_API_KEY (console: https://open.aminer.cn/open/board?tab=control). Never print the key. The scripts are pure stdlib — no dependency installation is needed except matplotlib for figure rendering (see requirements.txt).

Set the workspace once — the skill owns no ledger location, the path is the host's choice. The default (overridable via $DR_WORKDIR) is a per-run directory under the current project: outputs/<topic-slug>-<YYYYMMDD-HHMM>/, resolved to an absolute path at invocation time. Derive the slug from the topic (letters / digits / CJK / hyphens, ≤40 chars); every run gets its own directory, so runs never overwrite each other:

export DR_WORKDIR="${DR_WORKDIR:-$(pwd)/outputs/<topic-slug>-$(date +%Y%m%d-%H%M)}"
export DR_LEDGER="${DR_WORKDIR}/evidence-ledger.json"
mkdir -p "$DR_WORKDIR/figures"