The Agent Skills Directory

[SAFE]: The skill is primarily composed of instructional prompts and personality reference material. It does not contain scripts, binaries, or instructions for remote code execution.
[PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection as it processes user dialogue and utilizes the Read tool to ingest file content. However, the lack of dangerous capabilities (such as network access or shell execution) and the restriction of tools to Read and AskUserQuestion effectively mitigates the risk associated with this surface. Evidence chain:
Ingestion points: User dialogue and local files referenced via the Read tool.
Boundary markers: Not explicitly defined in the prompt instructions.
Capability inventory: Read (file access), AskUserQuestion (interaction). No command execution or network capabilities are present.
Sanitization: Relies on the underlying model's safety guardrails; no explicit sanitization is defined in the skill itself.

meadows