The Agent Skills Directory

[PROMPT_INJECTION]: The skill utilizes a pattern that is susceptible to indirect prompt injection by processing external task descriptions and GitHub issue data.
Ingestion points: Task data enters the workflow via /task <description> and /task #<issue> as defined in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions to the agent to disregard instructions potentially embedded within the task descriptions.
Capability inventory: The skill has the capability to perform file system operations, execute git commands (Steps 2 and 8), and trigger code generation and testing through subagents (Step 5).
Sanitization: No sanitization or input validation is specified for the external data ingested during the orientation or planning phases.
[COMMAND_EXECUTION]: The skill coordinates the use of version control system commands to manage the development environment.
Evidence: Step 2 (/isolate) and Step 8 (/ship) describe the use of git for operations such as worktree management, branch freshening, and merging.
[REMOTE_CODE_EXECUTION]: The skill facilitates the generation and execution of code as part of its primary software development purpose.
Evidence: Step 5 (/build) dispatches subagents for implementation and testing, which involves writing new code and executing test suites (TDD) based on the generated plans.

task