The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process untrusted documentation files (.md and .mdx) to identify jargon and generate simplified explanations. If these files contain malicious instructions embedded within the technical text, they could attempt to influence the agent's output or the subagent's verification process.
Ingestion points: Technical documentation files provided at runtime via the /eli5 command (defined in SKILL.md, Step 1).
Boundary markers: While the skill uses a structured 9-step workflow and specific content-type detection signals to categorize text, it lacks explicit delimiting or 'ignore instructions' warnings when passing documentation content to the LLM for analysis.
Capability inventory: The skill possesses capabilities for reading local files, generating text comparisons, and spawning subagents using the Task tool to verify claims against the repository's content. It does not execute code contained within the processed documents.
Sanitization: No explicit sanitization or filtering of the input documentation content is documented before it is processed by the language model.
Automated Subagent Verification: The skill uses a separate subagent to perform an 'adversarial review' of all net-new claims. This is a positive security practice that helps ensure technical accuracy and prevents the primary agent from hallucinating or being misled by confusing original text.

eli5