The Agent Skills Directory

[DATA_EXFILTRATION]: The skill includes instructions to read project context from a hardcoded local file path at /home/kidpixel/render_signal_server-main/memory-bank/activeContext.md. This path corresponds to the author's local environment and is used to initialize the agent's 'Memory Bank'.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of parsing and analyzing external documentation.
Ingestion points: The skill ingests untrusted data through the description and requirements fields of plan_task, as well as the summary and initialConcept fields of analyze_task.
Boundary markers: There are no defined boundary markers or explicit instructions to ignore embedded commands within the documents being processed.
Capability inventory: The skill has access to sensitive capabilities through integration with filesystem tools, specifically fast_read_file for reading context and fast_edit_block for implementing code changes based on task analysis.
Sanitization: No sanitization, validation, or escaping of the external input is specified in the skill's logic.

shrimp-task-manager