The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the gh (GitHub) CLI to perform repository and code searches. It explicitly instructs the agent to batch these operations into parallel Bash calls. While these are standard development tools, constructing shell commands from research queries presents a surface for command injection if inputs are not properly sanitized.
[EXTERNAL_DOWNLOADS]: The core functionality of the skill involves systematic retrieval of data from a wide variety of external sources, including technical forums (V2EX, Habr, Reddit), academic repositories (arXiv), and GitHub. It provides guidance on using numerous MCP servers and APIs for web scraping, crawling, and browser automation.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it fetches and processes untrusted data from the internet. The instructions require the agent to extract ideas and claims from these sources to inform subsequent research steps and the final report.
Ingestion points: Web content retrieved via WebFetch, gh search, and various retrieval MCP tools (Firecrawl, Exa, etc.).
Boundary markers: The skill does not provide specific instructions to use delimiters or safety markers when the agent processes retrieved text.
Capability inventory: Includes the ability to execute shell commands (gh CLI), write to the local file system (./report/), and access the network.
Sanitization: No specific sanitization or validation logic for the fetched content is described in the instructions.
[DATA_EXFILTRATION]: The skill aggregates information from diverse internet sources and writes a synthesized research report to the local file system in the ./report/ directory. While this is the intended research behavior, the process of collecting and centralizing data from many sources into local storage is a characteristic often monitored for data management and exfiltration risks.

deep-gather