The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes instructions to install the uv tool by downloading and executing a script directly from the well-known service astral.sh via a shell pipe.
[COMMAND_EXECUTION]: The skill employs Bash tools for system-level tasks, such as managing directory structures and registering agent commands by copying files into system directories.
[COMMAND_EXECUTION]: The skill implements a persistence mechanism by writing a Node.js monitoring script to the filesystem and modifying the agent's .claude/settings.local.json configuration to ensure the script executes during every tool call in future sessions.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external configuration files without explicit sanitization.
Ingestion points: The skill reads squad.yaml and .md agent/task files using the Read and Glob tools.
Boundary markers: The protocol does not specify the use of delimiters or instructions to ignore embedded commands within the ingested files.
Capability inventory: The skill has access to Bash, Write, and Edit tools to perform system-level modifications.
Sanitization: Contents are parsed and used to influence agent behavior without prior sanitization or validation of the input data.

squads