The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external content from files or directories to generate tickets, creating a surface for indirect prompt injection. However, the skill implements strong logical guardrails including a 'no-invention' rule, a 'stranger test', and requiring human-in-the-loop (HITL) approval for sensitive actions like creating issues. Ingestion points: The skill accepts a [file | dir | 'text'] as the primary source of truth for planning. Boundary markers: The instructions rely on logical constraints and 'stranger tests' rather than technical delimiters to separate source data from instructions. Capability inventory: The skill is authorized to write Markdown files and create issues on GitHub. Sanitization: No explicit content sanitization or escaping is mentioned for the generated ticket text.
[SAFE]: The skill includes explicit security instructions that mitigate common attack vectors. Specifically, it commands the agent to 'never read, list, grep, or count environment variables or files to find credentials' and requires explicit user authorization for repository access and issue creation.

hl-tickets