The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The skill frequently uses npx @graphql-inspector/cli and npm install -g @graphql-inspector/cli to fetch and execute packages from the public npm registry. These dependencies are not from the pre-approved trusted list.
COMMAND_EXECUTION (MEDIUM): The skill requires Bash access to execute CLI tools and shell scripts for auditing operations. This capability allows the agent to run arbitrary code downloaded from the internet.
INDIRECT_PROMPT_INJECTION (LOW): The skill processes untrusted local data that could contain malicious payloads.
Ingestion points: Reads contents from ./src/**/*.graphql, ./src/**/*.tsx, and ./schema.graphql.
Boundary markers: None identified; the tool parses file contents based on glob patterns without input sanitization.
Capability inventory: Access to Bash (via npx), Read (local files), and Write (audit reports).
Sanitization: Absent; the skill relies on the third-party @graphql-inspector parser to handle potentially malicious or malformed GraphQL documents.

graphql-inspector-audit