The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches task metadata and descriptions from an external source and interpolates them directly into a sub-agent prompt, creating a potential injection surface where malicious task content could influence the sub-agent's behavior.
Ingestion points: Task ID, title, and body are retrieved via agkan task get in SKILL.md.
Boundary markers: The instructions use Markdown headers for structure but do not include explicit instructions for the sub-agent to ignore commands or instructions embedded within the task body.
Capability inventory: The sub-agent is granted access to powerful CLI tools including git, gh (GitHub CLI), and agkan, along with the ability to modify the local filesystem.
Sanitization: There is no evidence of content validation or escaping of the task data before it is included in the sub-agent's prompt.
[COMMAND_EXECUTION]: The skill utilizes several local command-line tools to automate git operations and repository management.
Evidence: Extensive use of git (for status, stash, checkout, pull, log), gh (for repository and pull request management), and the agkan CLI tool for task lifecycle management.

agkan-run