The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the bash tool to interact with the alex binary. It supports both a pre-installed binary and running directly from source using go run ./cmd/alex. These commands manage team execution, status tracking, and tmux-based terminal interactions.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it accepts arbitrary strings for flags like --goal, --prompt, and --message and interpolates them directly into bash commands. This could allow maliciously crafted input from external data sources to influence the team's execution logic.
[PROMPT_INJECTION] Evidence Chain:
Ingestion points: Untrusted strings are ingested via the --goal, --prompt, and --message flags in SKILL.md.
Boundary markers: No boundary markers or delimiters are used to wrap the interpolated inputs.
Capability inventory: The skill uses the bash tool to perform command execution.
Sanitization: No input validation or sanitization routines are defined in the skill for these parameters.

team-cli