Virtual Reading Group
PassAudited by ClawScan on May 1, 2026.
Overview
This instruction-only skill is coherent for an academic reading group, but it will read user-selected papers, share generated notes among sub-agents, and write persistent output files.
This appears safe for its stated purpose. Before using it, choose a dedicated output folder, avoid mixing unrelated private files into that folder, and review generated notes/citations if the papers are sensitive or the conclusions matter.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Files placed in the chosen output directory may be read and summarized by the skill.
The synthesis phase is instructed to read the output directory. This is purpose-aligned, but if the user chooses a directory containing unrelated files, those files could be pulled into the synthesis context.
Read ALL files in `{output_dir}`Use a dedicated, empty project output folder for each reading group.
Paper contents and intermediate analyses may be visible across the spawned expert, junior, and synthesis agents.
The workflow intentionally passes notes, summaries, and discussion documents between multiple spawned agents. This is disclosed and necessary for the multi-agent reading-group function.
Phase 2 ... All Phase 1 outputs ... Phase 3 ... Phase 2 output + other experts' summaries
Avoid using sensitive or confidential papers unless you are comfortable with their contents being processed across the full multi-agent workflow.
Incorrect citations, hallucinated summaries, or adversarial text in source papers could propagate into later outputs.
Later phases rely on generated notes and summaries from earlier phases, so mistakes or misleading content in intermediate files can influence the final synthesis.
Read ALL files from Phases 1-3
Review intermediate notes and verify citations before relying on the final integrated summary.