The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external sources, creating a surface for indirect prompt injection attacks.
Ingestion points: The agent ingests data from the [Amplitude] Error Logged event, specifically the Error Message and Error Stack Trace properties, as well as the [Amplitude] Message property from [Amplitude] Error Click events via the Amplitude:query_dataset tool.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat these specific data fields as untrusted or to ignore any embedded commands within them.
Capability inventory: The skill possesses capabilities to query datasets, retrieve project context, and list deployments using the Amplitude toolset.
Sanitization: The instructions do not specify any sanitization, filtering, or escaping of the error-related strings before they are incorporated into the final report narrative in Phase 6. This could allow an attacker to influence the agent's summary or recommendations by triggering specific error messages containing malicious instructions.

monitor-reliability