The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates an orchestration workflow where a lead agent ingests and synthesizes output from multiple autonomous sub-agents ('teammates').
Ingestion points: SKILL.md instructions define a process for the lead agent to 'Collect all deliverables from output files' produced by teammates.
Boundary markers: The skill does not define specific delimiters or instructions to prevent the lead agent from obeying instructions potentially embedded in teammate output.
Capability inventory: The skill notes that spawned teammates possess full agent capabilities, including reading/writing files and executing commands, which could be leveraged if a sub-agent is influenced by malicious data.
Sanitization: There are no instructions for sanitizing, escaping, or validating the content generated by teammates before it is processed and summarized by the lead agent.
[SAFE]: The skill explicitly forbids the use of bash, subprocess, or CLI commands for agent orchestration, favoring native platform features and reducing the risk of shell-based command injection during team setup.
[SAFE]: No hardcoded credentials, sensitive file access (e.g., .ssh, .aws), or external network connections to untrusted domains were identified in the instructions or examples.

agent-teams