The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands such as sed -i, grep, find, and mkdir to perform file operations and update task statuses. These commands are triggered by natural language inputs to manipulate local documentation files.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted task data and user-provided variables.
Ingestion points: The agent reads task titles, descriptions, and implementation steps from Markdown files stored in docs/tasks/active/ using cat and grep.
Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to ignore potentially malicious content embedded within the task files it processes.
Capability inventory: The skill possesses the ability to execute shell commands, specifically using sed -i for in-place file editing and find for file searching, which can be exploited if variables are manipulated.
Sanitization: There is no evidence of sanitization or escaping of user-provided inputs (such as the task title or step numbers) before they are used as variables in shell scripts, creating a surface for command injection.

task-tracker