The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to format a user-supplied date string using date -d "${date}". Untrusted input from the --date argument is interpolated directly into a shell command string. While the allowed-tools configuration attempts to restrict execution to the date binary, the lack of input sanitization could allow for argument injection or the execution of subcommands if the shell environment allows it.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its data processing workflow.
Ingestion points: The skill ingests untrusted external data via the WebSearch and WebFetch tools (referenced through the research-company invocation) and reads local files like SUMMARY.md which contain external research.
Boundary markers: There are no instructions to use delimiters, XML tags, or "ignore embedded instructions" warnings when the agent processes the retrieved research content or the SUMMARY.md file.
Capability inventory: The skill possesses significant capabilities, including the ability to Write files to the filesystem and execute Bash commands (mkdir, test, date).
Sanitization: The procedure lacks any specified sanitization, validation, or filtering of the external content before it is used to synthesize talking points and questions, potentially allowing malicious content in the research data to influence the agent's behavior.

meeting-prep