The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages the execution of the claude CLI and its own background scheduler daemon.
Uses posix_spawn to launch agents with specific prompts.
Provides commands (--daemon, --wake, --stop) to control child process execution and monitoring.
[EXTERNAL_DOWNLOADS]: The skill references and utilizes resources associated with its author.
Built around the @forwardimpact/basecamp package.
Downloads and merges knowledge base templates and agent skills via the --update command.
[DATA_EXFILTRATION]: The skill accesses sensitive local data to populate the agent's knowledge base.
Reads Apple Mail threads and Calendar events from ~/.cache/fit/basecamp/.
Accesses local configuration and state files in ~/.fit/basecamp/.
This data is used locally to provide context to agents as part of the core functionality.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection via the data it processes.
Ingestion points: Content is ingested from external email threads and calendar events (found in ~/.cache/fit/basecamp/).
Boundary markers: None identified for isolating untrusted mail/calendar content from agent instructions.
Capability inventory: The skill can execute shell commands via the claude CLI using posix_spawn.
Sanitization: No evidence of sanitization or escaping of the ingested email/calendar data before it is interpolated into agent prompts.

fit-basecamp