The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands as part of its standard operation. This includes running a setup script (resolve_customization.py) using Python 3 and using system tools like grep and git log to analyze the codebase and project history during the investigation process.
[REMOTE_CODE_EXECUTION]: The skill provides a mechanism to execute arbitrary shell commands defined within the customize.toml configuration file. The activation_steps_prepend, activation_steps_append, and on_complete fields can be used to run any command on the host system. This represents a risk if the configuration file is modified by an untrusted source or included in a compromised repository.
[PROMPT_INJECTION]: The skill possesses a high surface area for indirect prompt injection because it is specifically designed to ingest and interpret data from untrusted sources.
Ingestion points: Untrusted data enters the agent context from ticket details (Outcome 1), diagnostic archives, log files, and stack traces (Outcome 2).
Boundary markers: There are no instructions to use delimiters or ignore potential commands embedded within the data being analyzed.
Capability inventory: The skill has the ability to execute shell commands, run local Python scripts, and perform parallel tool calls across the filesystem.
Sanitization: The instructions do not define any sanitization or filtering logic for the data retrieved from external logs or issue trackers before it is processed by the agent.

bmad-investigate