The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands (curl, pace) to fetch data. It explicitly suggests using user-provided inputs (e.g., time windows, panel IDs) as command-line flags or URL query parameters. This creates a potential surface for command injection if the agent does not properly sanitize user inputs before passing them to the shell.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources that are aggregated by the Pace service.
Ingestion points: Content is retrieved from localhost:7453/brief.md or the pace brief CLI command (described in SKILL.md).
Boundary markers: The instructions lack specific guidance for the agent to use delimiters or "ignore instructions" wrappers when processing the brief's content.
Capability inventory: The skill uses local command execution (curl, pace CLI) and is instructed to use web search tools for "named gaps."
Sanitization: There is no evidence of sanitization or filtering of the external content retrieved from the Pace brief before it is summarized for the user.
[DATA_EXPOSURE]: The skill accesses a local service at localhost:7453. While this is the intended functionality for a local brief service, it involves interacting with the local network environment.

pace-brief