The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands (ls, head, bd) to identify recent handoff files and check task status. These commands are restricted to the local environment and do not interact with remote resources or accept unsanitized user input into critical command segments.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and acts upon instructions contained in external Markdown files. Specifically, it is instructed to parse the 'Next Steps' from handoff documents and continue work based on those instructions.
Ingestion points: Processes external files within the thoughts/shared/handoffs/ directory (SKILL.md).
Boundary markers: No explicit delimiters or system instructions are provided to the agent to treat the handoff content as data rather than instructions.
Capability inventory: The skill can execute shell commands (ls, bd), interact with the TaskList tool, and perform file read operations.
Sanitization: There is no evidence of validation or sanitization of the content extracted from the handoff files before it influences the agent's next actions.

restore