The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates a bash script to poll an inbox and schedules it for execution.
Evidence: Writes to /data/workspace/scripts/squadron-inbox-poll.sh and uses schedule_task to run the script every 30 minutes.
[EXTERNAL_DOWNLOADS]: The skill retrieves and sends data to a remote service.
Evidence: Connects to https://community.iamstarchild.com/1892-squadron/api to sync tasks and knowledge base files.
[SAFE]: Sensitive credentials are managed through standard environment configurations.
Evidence: Instructions prioritize SQUADRON_API_KEY from the environment and suggest storing secrets in a .env file.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from data retrieved via the Squadron platform.
Ingestion points: External data enters through the /inbox and /tasks API endpoints defined in the skill.
Boundary markers: Absent; the skill does not wrap external content in delimiters or specify that instructions within the data should be ignored.
Capability inventory: The skill has access to scheduled shell execution, file writing to the knowledge base, and network connectivity.
Sanitization: Absent; task titles and descriptions are processed and displayed without escaping or validation.

@1892/squadron