Scholar Deep Research

Use when the user asks for a literature review, academic deep dive, research report, state-of-the-art survey, topic scoping, comparative analysis of methods/...

This skill appears to implement a legitimate multi-source literature-review workflow, but it performs two behaviors you should review before installing: 1) Automatic update: Phase 0 runs scripts/check_update.py which does a git fetch and can fast-forward the skill repository before or during a run. That means code executed by your agent could change when the skill auto-updates. If you want a fixed codebase, set SCHOLAR_SKIP_UPDATE_CHECK=1 (documented in SKILL.md) and pin the repo to a commit. 2) External script execution and undeclared env-vars: The code may shell out to helper scripts (e.g., paper-fetch) discovered at configurable paths (~/.claude/skills/paper-fetch/scripts/fetch.py or PAPER_FETCH_SCRIPT). SKILL.md also references env vars (SCHOLAR_SKIP_UPDATE_CHECK, SCHOLAR_MAILTO, PAPER_FETCH_SCRIPT) even though the skill metadata lists none. Before running, inspect the repository locally (especially scripts/check_update.py and scripts/extract_pdf.py) to confirm exactly what network calls and subprocess executions they make. Practical recommendations: - Review the scripts in the repo locally (check_update.py, extract_pdf.py, and any subprocess calls) before installing. - Run the skill in an isolated environment (container or VM) the first time, with SCHOLAR_SKIP_UPDATE_CHECK=1 to avoid auto-updates until you're comfortable. - If you do allow updates, monitor the repo's upstream history and consider installing from a release tag rather than the branch HEAD. - Be aware the skill will write research_state.json and .last_update_check to the skill directory; ensure the install path has appropriate permissions and is isolated from sensitive data. Given these surprises (auto-updating code, undeclared env vars, and optional shelling to other local scripts), treat the skill as useful but with elevated operational risk unless you pin or audit it first.

Type: OpenClaw Skill Name: scholar-deep-research Version: 0.5.0 The skill bundle implements a legitimate academic research workflow but contains high-risk behaviors that warrant caution. Specifically, `check_update.py` performs an automated `git pull` to update the skill's own code from its repository, which introduces a supply-chain risk via remote code modification. Additionally, `extract_pdf.py` uses `subprocess.run` to execute an external script (`fetch.py`) discovered by searching through various hidden directories associated with other AI agent skills (e.g., `~/.claude/skills/paper-fetch/`). While these features are documented and support the stated goals of maintenance and interoperability, they represent significant attack surfaces for lateral execution and unauthorized code updates.