Imbad0202/academic-research-skills is an open-source suite of Claude Code skills designed to augment the academic research pipeline through a human-in-the-loop framework. Rather than automating authorship, the tool addresses systemic failure modes identified in autonomous research systems—such as citation hallucination, methodology fabrication, and "frame-lock"—by implementing mandatory integrity gates at stages 2.5 and 4.5. The system utilizes a multi-agent architecture, including a 13-agent research team and a 12-agent writing pipeline, to handle tasks like Socratic-guided planning, literature reviews, and LaTeX formatting. To combat AI-specific risks like sycophancy and premature convergence, the suite incorporates advanced mechanisms such as a Concession Threshold Protocol for its "Devil's Advocate" agent and an Intent Detection Layer for its "Socratic Mentor." Technical robustness is reinforced through "trust-chain" frontmatter for source provenance and locator infrastructure (three-layer citation anchors) introduced in v3.7.3 to enable claim-level audits. The tool's design is motivated by empirical findings, including Zhao et al.'s (2026) audit of 111M references which identified significant hallucinated citation trends, and studies showing that AI-led verification performs significantly worse (37% reproduction rate) than human-only (94%) or AI-assisted (91%) teams.