The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data sources such as logs and system state, creating a theoretical surface for indirect prompt injection.
Ingestion points: SKILL.md specifies logs, system state, code, and documentation as diagnostic inputs.
Boundary markers: The instructions include explicit directives to deny harmful or security-compromising actions.
Capability inventory: The skill restricts itself to evidence gathering; mutation is explicitly forbidden via the remediate: no and deny: target mutation flags.
Sanitization: No specific data sanitization is described within the instruction set.

dig-deeper