The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The setup documentation and configuration files guide the user to download packages from the vendor's repository and registry. This includes cloning the awslabs/mcp repository and using uvx to fetch the cloudwatch-mcp-server and cloudwatch-applicationsignals-mcp-server packages.
[REMOTE_CODE_EXECUTION]: The skill employs uvx to download and run the CloudWatch and Application Signals MCP servers provided by the vendor. This is the intended mechanism for enabling CloudWatch integration within the agent environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process log data from external sources (CloudWatch Logs), which represents an attack surface for indirect prompt injection.
Ingestion points: Untrusted log data enters the agent context through the get_logs_insight_query_results tool as described in SKILL.md.
Boundary markers: The skill provides structured query templates and field extraction logic, though it does not explicitly define markers to encapsulate the log content from the agent's instructions.
Capability inventory: The associated agent configuration includes capabilities for file system operations and shell command execution as documented in kiro-skill-setup.md.
Sanitization: The skill implements specific noise filtering heuristics and field prioritization to selectively process telemetry data and ignore infrastructure boilerplate as detailed in references/otel-span-schema.md.

agentcore-investigation