Chalk: Fact-Checked Tutor Mode

Technical educators and students often struggle with LLMs that prioritize agreeability over accuracy. Chalk is a pedagogical skill that replaces passive explanation with a rigorous teach-and-test loop designed to catch both student misconceptions and AI hallucinations. It transforms a standard chat into a structured tutoring environment that enforces active recall and cross-references claims against external ground truth.

What it does

• Active Recall Enforcement delivers small units of information followed by mandatory quizzes to ensure retention before moving forward.
• Anti-Hallucination Guardrails implement a two-tier fact-checking system, using subagents to verify high-stakes claims like Big-O complexity or formulas against real-world sources.
• Pedagogical Linting scans explanations for teaching-order bugs, ensuring all terms are defined and prerequisites are met before introducing complex topics.
• Structural Visualization generates ASCII or Mermaid diagrams for data structures, recursion call stacks, and pointer logic to provide mental models.
• Session Continuity maintains a detailed chalk-log.md to track quiz scores, open bugs in understanding, and specific "pick-up" points for future sessions.

How it works

1. Initialize by typing /chalk or "teach me [topic]" to trigger the learning loop and load your style.md preferences.
2. Execute Loop by consuming one concept at a time, completing a mandatory "explain-it-back" gate or quiz before the next module unlocks.
3. Verify Claims by requesting /chalk check <claim> to trigger a subagent retrieval that attempts to refute the stated fact with external sources.
4. Sync Vault by pointing the agent to an Obsidian or local directory to incorporate existing notes into the context while flagging contradictions in those notes.

Frameworks & tools

This skill is optimized for Obsidian vaults and local Markdown environments. It uses Mermaid for diagramming and requires an agent capable of tool-use for web retrieval and subagent spawning to handle high-stakes fact-checking.

Why this beats prompting it yourself

Standard prompts often fail to catch their own errors and tend to "dump" information rather than teaching it. Chalk forces the agent into a disciplined pedagogical structure that identifies the root cause of a student's mistake rather than just correcting the surface slip.

Use cases

• Master complex Data Structures and Algorithms with automated Big-O verification.
• Audit existing study notes for technical inaccuracies or missing prerequisites.
• Prepare for technical interviews using a spaced-repetition loop that targets weak spots.

Known limitations

Obscure facts without stable web sources may still result in plausible but incorrect claims. Tier 2 fact-checking requires an environment that supports subagent delegation and web access.