The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands using the agkan CLI tool and jq for configuration management and task retrieval.
Evidence: CONFIG=$(agkan config get --json 2>/dev/null || echo '{}') and agkan task list --status icebox --json in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves untrusted data from task titles and bodies and interpolates them directly into a sub-agent prompt, creating a surface for indirect prompt injection attacks.
Ingestion points: Task data is fetched via agkan task list and agkan task get in SKILL.md (Steps 1 and 2).
Boundary markers: The prompt uses basic Markdown headers (## Task Information) but lacks explicit delimiters or instructions for the sub-agent to ignore potential commands embedded within the task body.
Capability inventory: The workflow involves shell command execution via the agkan CLI and delegation to sub-agents which follow procedures in agkan-icebox-subtask-SKILL.md.
Sanitization: There is no evidence of sanitization, escaping, or validation of the interpolated <body> or <title> variables before they are passed to the sub-agent prompt.

agkan-icebox