The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external work graph, creating an attack surface where malicious instructions embedded in task titles or comments could influence agent behavior.
Ingestion points: The skill retrieves external content using bd show <epic>, bd children, and bd comments <bead> as described in the 'Observe' section of SKILL.md.
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following natural language instructions found within the bead data.
Capability inventory: The skill has the ability to create and modify the graph using bd create and bd update, and can delegate tasks to other skills like Orchestrate or Plan.
Sanitization: The instructions do not specify any sanitization, filtering, or validation of the content retrieved from the 'bd' tool before it is used to determine the next 'wave' of work.

navigate