The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill acts on external data from Basecamp events and project files, which constitutes a surface for indirect prompt injection.
Ingestion points: Processes user-provided message content (content) from Basecamp and configuration instructions from repository-hosted AGENTS.md files (referenced in 'Choose the repo' and subagent sections).
Boundary markers: Does not implement explicit delimiters or 'ignore embedded instructions' warnings for the external content.
Capability inventory: Subagents are granted shell access to perform repository operations, file modifications, and Basecamp API interactions via the CLI.
Sanitization: The skill removes the agent's own mention but does not describe further sanitization or escaping of the incoming request body.
[COMMAND_EXECUTION]: The skill frequently executes the basecamp CLI tool and orchestrates background tasks. It includes robust security rules to prevent command injection, such as mandatory single-quoting for all variable inputs and specific escaping for nested single quotes.

basecamp-connect