local-file-grounding

What problem does it solve?

It prevents hallucinations by forcing responses to be derived from the task’s local directory files rather than relying on generic knowledge.

Core Features & Use Cases

• File-first reasoning: Inspect the task directory, identify required inputs/outputs, and read the named files before concluding.
• Grounded summarization and quoting: Summarize or quote only what the files support, keeping claims aligned to the source material.
• Deterministic output writing: When output is required, create or update the requested result file in the task directory in a way that matches the read contents.
• Use case: You have an analysis request that references a specific local CSV and a JSON spec; the Skill reads those files and produces a grounded report plus a properly written output artifact.

Quick Start

Use this skill to generate a grounded summary and write an output file based on the local task directory contents.