The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external source that could contain malicious instructions.
Ingestion points: Task titles and priority values are retrieved via the agkan tool in SKILL.md.
Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when displaying task information.
Capability inventory: The skill uses bash to execute agkan and jq for data processing.
Sanitization: Task content is displayed without sanitization or escaping of potentially harmful prompt-altering text.
[COMMAND_EXECUTION]: The skill executes local shell commands to perform its primary function.
Evidence: Invokes agkan task count, agkan task list, and jq within the workflow steps defined in SKILL.md.

dashboard