The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate a workflow where the agent ingests untrusted data, creating a potential surface for indirect prompt injection.
Ingestion points: SKILL.md directs the agent to gather evidence from potentially untrusted sources including code, logs, traces, and database plans.
Boundary markers: The instructions do not mandate the use of delimiters or specific system instructions to treat gathered evidence as data rather than instructions.
Capability inventory: The skill is used in a context where the agent is empowered to implement and verify code fixes, which implies file write access.
Sanitization: There are no mentions of sanitizing or escaping the evidence collected during the debugging process.

effect-typescript-debugging