The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes a workflow where agents ingest and process data from external sources, which constitutes a vulnerability surface.
Ingestion points: Agents are instructed to use kanban_show to read the summary, metadata, and comments from previous task runs. The task body itself and the unblock comments are also primary inputs.
Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands within the ingested task data.
Capability inventory: The agent has access to the local file system (via $HERMES_KANBAN_WORKSPACE), shell execution (git commands), and task management tools (kanban_create, kanban_complete, kanban_block).
Sanitization: There is no mention of sanitizing or validating the content of the summary, metadata, or comments before processing or interpolating them into future actions.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands, specifically git worktree add, using environment variables such as $HERMES_KANBAN_WORKSPACE, $HERMES_KANBAN_BRANCH, and $HERMES_KANBAN_TASK. It also describes the use of the hermes CLI for task lifecycle management.

kanban-worker