Recursive strategies to let Claude Code tackle massive inputs beyond standard context limits.
The Recursive Decomposition Skill equips Claude Code with structured methods to manage inputs that would normally exceed practical context sizes. Rather than loading entire repositories or document sets at once, it teaches the model to narrow scope first, partition work strategically, and delegate segments to focused sub-processes before synthesizing results.
Core mechanics include upfront filtering to shrink search space, deliberate chunking of material, spawning independent analysis threads, targeted re-verification on smaller windows, and programmatic aggregation of outputs. This mirrors map-reduce patterns while staying inside Claude Code's plugin system.
It targets developers and analysts who regularly work with dozens or hundreds of files, multi-document QA, or codebase-wide searches where standard context windows cause degraded accuracy or incomplete coverage.
Scan hundreds of files to locate API usage patterns, error handling approaches, or security issues while returning precise file and line references.
Pull and combine key facts from large collections of documents or meeting notes without losing details that fall outside normal context windows.
Process entire repositories or document sets systematically by breaking them into manageable segments that sub-agents can handle in parallel.
claude plugin install recursive-decomposition@recursive-decomposition# Add the marketplace
claude plugin marketplace add massimodeluisa/recursive-decomposition-skill
# Install the plugin
claude plugin install recursive-decomposition@recursive-decompositionIt is designed for tasks involving 10 or more files or roughly 50k+ tokens where standard context handling begins to degrade.
Verified reviews from the community shape this listing's rating.
Loading reviews…