The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines a workflow for investigating security breaches and operational incidents that relies on the ingestion of untrusted data, creating an attack surface for Indirect Prompt Injection.
Ingestion points: The documentation in SKILL.md requires the agent to read and analyze 'log files', 'system snapshots', 'network captures', and 'user activity'.
Boundary markers: The instructions do not define any delimiters or system-level warnings to distinguish between trusted instructions and potentially untrusted content within the analyzed files.
Capability inventory: The skill instructs the agent to exercise high-impact capabilities including 'access revocation', 'traffic blocking', 'process termination', and 'system shutdown' based on its findings.
Sanitization: There are no instructions provided to sanitize, validate, or verify the integrity of the data being analyzed before it influences agent decision-making.
[DATA_EXFILTRATION]: The skill requires access to highly sensitive system artifacts, such as 'memory dumps', 'system snapshots', and 'log preservation'. While no specific external transmission URLs are provided, the collection of this data represents a significant data exposure risk if the agent is compromised.
[COMMAND_EXECUTION]: The workflow requires the agent to implement solutions and perform containment actions that involve executing critical system commands. Although no commands are hardcoded in the skill files, the mandate to 'isolate services' and 'quarantine data' provides a path for unauthorized command execution if the agent's logic is subverted.

incident-responder