The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific documentation files (CONTEXT.md, GLOSSARY.md, UBIQUITOUS_LANGUAGE.md, and *.glossary.yml) to align terminology, representing an ingestion point for data that could potentially contain malicious instructions. Ingestion points: Files matching the project documentation patterns specified in the procedure. Boundary markers: The skill contains explicit constraints instructing the agent to 'not perform a new task' and 'not introduce information from the terminology files that the previous response did not contain.' Capability inventory: The skill is restricted to text re-explanation and does not possess capabilities for file writing, network operations, or shell command execution. Sanitization: Strict writing rules (inspired by ASD-STE100) and clarity controls limit the agent's behavior and output format, serving as a functional constraint.
[NO_CODE]: The skill consists entirely of markdown instructions and YAML configuration. It does not include any executable scripts, binaries, or dynamic code execution patterns, which significantly minimizes the attack surface.

wtf