The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requests broad execution permissions via allowed-tools: Bash(infsh *). This grants the agent the ability to execute any command or application available through the infsh CLI, which serves as an interface to various remote services and local utilities.
[REMOTE_CODE_EXECUTION]: The skill employs the infsh/python-executor tool to run Python code for generating benchmark charts. This involves generating and executing scripts at runtime, which is a form of dynamic execution. While the provided example is benign, the underlying capability allows for arbitrary code execution within the executor's environment.
[PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection by design, as it ingests and processes untrusted data from external sources.
Ingestion points: Search query results from the exa/search tool are used to research and write technical content.
Boundary markers: No explicit delimiters or instructions are provided to the agent to distinguish between the skill's instructions and potentially malicious content found in search results.
Capability inventory: The skill can execute subprocesses through infsh, run dynamic Python code, and post content to social media via x/post-create.
Sanitization: There is no evidence of filtering or sanitizing the data retrieved from external tools before it is processed by the agent.

technical-blog-writing