The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to poll and read external state from sources like continuous integration (CI) systems, deployment logs, or remote queues to decide whether to continue a loop. This creates an indirect prompt injection surface where malicious content in those external systems could influence the agent's behavior.
Ingestion points: External state sources such as CI systems, deployment logs, and remote queues (SKILL.md, Lever 2).
Boundary markers: Not specified in the instructions.
Capability inventory: The skill references capabilities including Bash, Agent, and Workflow (SKILL.md, Lever 2).
Sanitization: No sanitization or validation of external inputs is mentioned.
[COMMAND_EXECUTION]: The skill references using Bash and git rev-parse as mechanisms for short-circuiting work. These are described as standard tools for checking local state and do not involve untrusted input or risky execution patterns.

honey-loop