The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python helper script (manage_labels.py) that executes the GitHub CLI (gh) via the subprocess module.\n
Commands are constructed as lists (e.g., ["gh", "label", "list", ...]) which prevents shell injection.\n
The execution is limited to label management tasks such as listing, creating, and deleting labels.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external GitHub repositories, specifically label names and descriptions, which could potentially contain malicious instructions.\n
Ingestion points: Data enters the system via the list_labels function in manage_labels.py using gh label list.\n
Boundary markers: The SKILL.md instructions explicitly tell the agent to display existing labels in a separate, clearly titled section to distinguish them from the instructions.\n
Capability inventory: The skill has the capability to modify repository labels via the GitHub CLI.\n
Sanitization: The Python script parses the CLI output as JSON and treats the values as literal strings, comparing them against a hardcoded preset rather than allowing the agent to interpret them as instructions.\n- [SAFE]: The skill implements several industry-standard safety practices for destructive operations.\n
It requires explicit user approval before any changes are made.\n
It utilizes a Plan ID (SHA-256 hash of the state) to ensure that changes are only applied if the repository state has not changed since the user's approval.\n
It performs a verification step after execution to ensure the results match the approved plan.

init-ghp-labels