The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches feature descriptions and acceptance criteria from external services like Linear, GitHub, and Jira. These sources are potentially attacker-controlled and could contain malicious instructions designed to manipulate the agent's behavior during the implementation phase.
Ingestion points: The 'Create' subcommand in SKILL.md fetches details from issue tracker URLs or ticket IDs.
Boundary markers: There are no defined boundary markers or instructions for the agent to ignore embedded commands within the fetched ticket data.
Capability inventory: The skill utilizes the Agent, Task, Bash (restricted to git, ls, cat), Edit, and Write tools to perform code modifications and command execution.
Sanitization: The skill does not perform validation or sanitization on the external content before using it to generate implementation tasks.
[DYNAMIC_EXECUTION]: The skill follows a pattern of generating a structured plan in Markdown format and then executing the instructions contained within that plan using the Agent tool. This creates a dynamic execution flow where generated instructions drive subsequent agent operations.
Evidence: The next subcommand reads task specifications from the plans/ directory and uses them to instruct agents to implement code changes.

superplan