The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data from external sources like GitHub issues and pull requests, which may contain malicious instructions designed to manipulate agent behavior. * Ingestion points: Data enters the context via get-issue, search-prs, and repository file reads as described in SKILL.md. * Boundary markers: references/agentic-setup.md defines an 'Untrusted content boundary' that explicitly instructs the agent to treat external content as data and ignore any embedded commands or directives. * Capability inventory: The skill has the ability to execute shell commands (git), install packages, and mutate tracker state (labeling/commenting). * Sanitization: references/agentic-setup.md requires strict regex validation (^[A-Za-z0-9._/-]+$) for externally-sourced values and SKILL.md mandates redacting credentials from outputs.
[COMMAND_EXECUTION]: The skill performs shell-based operations to manage the development environment and project lifecycle. * Evidence: references/worktree-setup.md contains shell scripts for git worktree management and dependency installation using tools like npm, pnpm, or cargo. * Mitigation: The skill prevents command injection by sanitizing all variables (like issue IDs and slugs) before interpolation and ensures all destructive work occurs in isolated temporary directories.

om-auto-fix-issue