The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages a local practice environment by executing scripts (scripts/finish_problem.py, scripts/get_status.py) and Git commands. It also runs user-generated Python solutions to verify algorithm correctness during the learning process.
[DATA_EXPOSURE_AND_EXFILTRATION]: As part of its documented workflow, the skill automatically performs a git push to remote repositories to sync progress. While this is an intended feature for tracking goals, it involves automated network activity that transmits local repository state to a remote origin.
[PROMPT_INJECTION]: The skill instructions in SKILL.md use strict directives (e.g., "硬性规则", "严禁提示") to ensure the AI agent does not provide answers prematurely. These are instructional constraints designed to enhance the educational experience and do not attempt to bypass security filters.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external local sources such as leetcode.json and README.md (via a migration script).
Ingestion points: Reads leetcode.json to determine user settings/progress and parses README.md to import problem history.
Boundary markers: None identified in the logic that interpolates these files into the agent's context.
Capability inventory: The skill has the ability to run shell commands for Git and execute local Python scripts.
Sanitization: Content from these files is used directly without validation or escaping beyond basic parsing.
[METADATA_POISONING]: Documentation in references/progress_tracking.md mentions Todoist integration features (like searching tasks and adding comments) that are not implemented in the provided codebase. This appears to be a documentation error rather than a malicious trap.

leetcode-teacher