The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides extensive administrative control over AWS CloudWatch resources.
The PutResourcePolicy command allows the agent to modify access control policies. An example in the documentation shows a highly permissive policy configuration (Principal: "*").
Commands such as DeleteLogGroup and DeleteLogStream enable the permanent removal of log data.
PutSubscriptionFilter allows the agent to configure the delivery of log events to other AWS services or external endpoints.
[DATA_EXFILTRATION]: The skill facilitates the retrieval of potentially sensitive information through log inspection.
Commands like GetLogEvents, FilterLogEvents, and StartQuery allow the agent to read arbitrary log content which may contain secrets, PII, or system details.
When combined with commands like PutSubscriptionFilter or PutLogEvents, there is a risk that this data could be exfiltrated to unauthorized destinations.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted external data.
Ingestion points: Data enters the agent's context through log events retrieved via GetLogEvents, FilterLogEvents, and GetQueryResults (defined in SKILL.md).
Boundary markers: The instructions lack any boundary markers (e.g., XML tags, triple backticks) or specific guidance for the agent to ignore instructions embedded within log content.
Capability inventory: The agent has significant capabilities including resource deletion, policy modification, and log writing across multiple scripts.
Sanitization: There is no mention of sanitizing or validating the content of logs before the agent processes them.

awsclaw-cloudwatch