The Agent Skills Directory

COMMAND_EXECUTION (LOW): The skill recommends using bash -c to execute curl commands. This pattern is intended to preserve environment variables but introduces a potential shell injection surface if the agent populates placeholders (like your-org-name or team-name) with unvalidated data from external sources.
DATA_EXFILTRATION (SAFE): Network requests are directed exclusively to api.github.com, which is a trusted domain for this context. The primary function of the skill is to transmit data to and from this API.
CREDENTIALS_UNSAFE (SAFE): Authentication is handled via the GITHUB_TOKEN environment variable, which is correctly declared as a secret in the skill's YAML frontmatter. Documentation examples use standard placeholders (ghp_xxxxxxxxxxxx).
INDIRECT_PROMPT_INJECTION (LOW): The skill processes data from GitHub API responses.
Ingestion points: JSON responses from GitHub (e.g., list of seat assignments, usage metrics).
Boundary markers: No explicit boundary markers or delimiters are used in the command templates for data interpolation.
Capability inventory: File writing to /tmp and execution of shell commands via bash -c.
Sanitization: No explicit sanitization or validation of API-returned data before it is used in subsequent CLI operations is defined within the skill itself.

github-copilot