The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from GitHub repositories to create and triage issues, creating a surface for embedded instructions to influence agent behavior. * Ingestion points: The skill reads repository metadata, issue templates (YAML and Markdown), contribution guides (CONTRIBUTING.md), and existing issue lists/details via the GitHub CLI. * Boundary markers: Implements strict human authorization gates for all remote operations and specific host/repository destinations. It uses structured YAML form schemas to strictly map answers to declared controls rather than accepting arbitrary input. * Capability inventory: The skill possesses the capability to create issues, view issue details, and read repository labels using gh issue create, gh issue view, and gh api. * Sanitization: Requires a mandatory privacy scan and redaction of the final issue body immediately before publication to replace sensitive strings such as credentials, private hostnames, and home paths with placeholders.
[COMMAND_EXECUTION]: The skill executes local commands and interacts with remote APIs via the GitHub CLI (gh) and git. * Evidence: It uses gh auth status, gh repo view, gh issue list, and gh issue create to perform discovery and mutations. It also uses git commands for repository path resolution and mktemp for secure temporary file handling with umask 077.

gentle-ai-issue-creation