The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's search workflow pipes the output of a network request directly into a python3 interpreter. This pattern is a recognized attack vector, as a compromise of the remote server could lead to arbitrary code execution if the input is not strictly controlled.
[EXTERNAL_DOWNLOADS]: Fetches search results and documentation content from context7.com using curl.
[DYNAMIC_EXECUTION]: Utilizes python3 -c to execute an inline script defined in the skill instructions to parse JSON data.
[COMMAND_EXECUTION]: Relies on system commands (curl, python3) to perform its primary documentation retrieval tasks.
[REMOTE_CODE_EXECUTION]: The documentation suggests that users or agents install an MCP server using npx -y @upstash/context7-mcp, which downloads and executes remote code from the npm registry at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted documentation content from an external API and presents it to the agent, creating a surface for indirect prompt injection.
Ingestion points: External API responses from context7.com containing documentation and search results (SKILL.md).
Boundary markers: No delimiters or instructions are used to separate the external documentation content from the agent's primary instructions.
Capability inventory: The skill environment allows for shell command execution (curl, python3) which could be abused if an injection is successful.
Sanitization: There is no evidence of sanitization or validation of the content fetched from the remote API.

dotnet-library-docs