The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from existing GitHub issues and comments, which represents a potential attack surface for indirect prompt injection from external contributors.
Ingestion points: The agent reads external issue metadata and comments to synchronize the ORCHESTRATOR-ROADMAP.md and manage the project lifecycle (SKILL.md).
Boundary markers: The skill contains a robust "Untrusted Input Handling" section that explicitly warns the agent to treat external content as data, never as instructions, and to ignore embedded directives such as "ignore previous instructions."
Capability inventory: The skill has the ability to execute shell commands via the GitHub CLI (gh), including creating/modifying issues, labels, and project items.
Sanitization: The instructions mandate that any detected directives in untrusted input must be quoted verbatim to the user for approval rather than being executed automatically.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands and automation logic to interact with the GitHub environment.
Evidence: The skill provides instructions for executing gh CLI commands for authentication, repository management, and issue tracking.
Dynamic components: It describes a bulk creation process using a shell loop (while loop reading from slices.txt) which uses variables for both issue titles and file paths (slices/${title}.md). While intended for use with approved local sources, this pattern requires the agent to ensure that input strings are properly handled to prevent unintended shell behavior or path traversal.

create-issues