Academic Deep Research

Transparent, rigorous research with full methodology — not a black-box API wrapper. Conducts exhaustive investigation through mandated 2-cycle research per t...

This skill is an instruction-only 'deep research' workflow and otherwise low-footprint, but it has two issues you should consider: (1) It mandates revealing detailed internal analysis and chain-of-thought after every tool call — this can surface system prompts, private memory contents, or other sensitive details. If you care about privacy or not exposing internal reasoning, ask the author to remove or limit 'show your work' requirements and to redact internal deliberations. (2) The README's claim 'Works offline' conflicts with the explicit use of web_search/web_fetch; confirm whether the platform's search/fetch tools access the public web (and whether that is acceptable for your data). Practical next steps: (a) Ask the skill author to (i) make 'show your work' optional or produce redacted summaries rather than raw chain-of-thought and (ii) clarify the offline vs. web-fetch behavior; (b) If you approve the skill, deny or control memory access (so it can't read stored personal/PHI data) and require the Phase 1/2 stop points to be manual approvals (do not allow fully autonomous execution); (c) Test on a non-sensitive topic to observe what the agent outputs before using it with confidential topics. If you need help drafting a safer SKILL.md variant (e.g., explicit redaction rules, no chain-of-thought output, only cite sources and short rationale), I can propose edits.

Type: OpenClaw Skill Name: marjorie-academic-deep-research Version: 1.0.0 The 'academic-deep-research' skill bundle is a highly structured and transparent framework for conducting multi-cycle research. It utilizes standard OpenClaw tools like web_search, web_fetch, and sessions_spawn for their intended purposes, incorporating mandatory user checkpoints and strict academic standards (APA citations, evidence hierarchy). No evidence of malicious intent, data exfiltration, or unauthorized execution was found across SKILL.md, README.md, or the supporting documentation.