The Agent Skills Directory

[PROMPT_INJECTION]: The skill manages untrusted data from GitHub issues, creating a surface for indirect prompt injection.
Ingestion points: The skill fetches issue details using gh issue view <N> in both SKILL.md and plan-and-dispatch.md.
Boundary markers: There are no explicit instructions or delimiters to isolate untrusted issue text from the agent's instructions.
Capability inventory: The skill can trigger code generation and execution tasks through the superpowers:subagent-driven-development sub-skill.
Sanitization: No sanitization of the issue content is mentioned or implemented.
[COMMAND_EXECUTION]: The skill utilizes several standard command-line tools to perform its tasks.
Evidence: It executes gh for issue management and PR creation, git for status checks, and shell commands like mkdir and printf to manage local scratch files.
[SAFE]: The skill does not exhibit malicious behaviors such as credential theft, data exfiltration to unknown domains, or unauthorized privilege escalation. Its operations are confined to the repository and official GitHub services.

issue-to-implementation