The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill attempts to fetch a ROADMAP.md file from the vendor's GitHub repository at https://raw.githubusercontent.com/rohitg00/ai-engineering-from-scratch/main/ROADMAP.md if a local version is not present. This is a documented resource provided by the skill author for curriculum metadata.
[INDIRECT_PROMPT_INJECTION]: The skill collects free-text responses from the user and fetches remote data, then writes them into a LEARNING.md file. This file serves as the source of truth for subsequent skills (e.g., learn), creating a potential pathway for malicious instructions to be persisted into the agent's long-term context.
Ingestion points: User responses to interview questions (Step 1) and the contents of the externally fetched ROADMAP.md (Step 3).
Boundary markers: The skill uses Markdown headers and comments as delimiters in the generated file, but these are for structural organization rather than security sandboxing.
Capability inventory: Includes reading and writing to the local file system and making network requests to GitHub.
Sanitization: The instructions do not specify any validation, escaping, or sanitization of user input before it is written to the persistent state file.

start-learning