The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from GitHub issues and commit messages to orchestrate tasks, creating a surface for indirect prompt injection. Malicious content in an issue body or title could potentially influence the orchestrator's behavior or be propagated into shell commands. * Ingestion points: The skill uses gh issue view to read specification details, sub-issue descriptions, and blockers, and evaluates the "done set" based on Git commit messages. * Boundary markers: The instructions lack explicit delimiters or warnings to treat external data as untrusted, which could lead to instructions in an issue being interpreted as commands by the agent. * Capability inventory: The skill possesses significant capabilities, including creating new GitHub issues via gh issue create, managing Git branches and worktrees, and invoking subagents for code generation and review. * Sanitization: No sanitization is performed on variables such as issue titles or body text before they are used to brief subagents or interpolated into shell commands.

implement-spec