The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands through tools like agkan, git, and gh to manage task statuses, repository branches, and pull request creation. These are standard operations for the intended workflow.
[DATA_EXFILTRATION]: Includes defensive instructions specifically advising against using git add -A or git add . to prevent the accidental inclusion of sensitive files such as .env or credentials.* in version control commits.
[PROMPT_INJECTION]: Processes external content from task bodies and metadata using agkan task get. This creates a surface for indirect prompt injection where malicious instructions embedded in a task could attempt to influence the agent's implementation steps.
Ingestion points: Task content and metadata retrieved via agkan task get <id> --json.
Boundary markers: The workflow does not define specific delimiters to isolate external task instructions from the agent's system instructions.
Capability inventory: The agent has capabilities to modify the local filesystem, push to remote repositories, and create pull requests.
Sanitization: There is no explicit sanitization step for the task body before it is used to guide the implementation phase.

agkan-subtask