The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external token metadata such as name, symbol, and logo values which are set by token deployers and can contain malicious instructions.
Ingestion points: Metadata fields retrieved via gmgn-cli token info, security, trenches, and trending commands.
Boundary markers: The skill documentation notes that gmgn-cli sanitizes output by stripping control, zero-width, and bidi characters, and replaces suspicious text with the literal [filtered] string.
Capability inventory: The skill executes shell commands via gmgn-cli and python3 but limits operations to read-only API calls and data filtering.
Sanitization: Relies on gmgn-cli built-in sanitization and provides explicit instructions to the AI agent to treat string values as data only and ignore any embedded instructions.
[DYNAMIC_EXECUTION]: The skill uses python3 -c to execute localized scripts for filtering and parsing JSON data received from gmgn-cli command pipelines (found in Step 5B).
[COMMAND_EXECUTION]: The skill executes multiple sub-commands of gmgn-cli and utility commands like ifconfig or ip addr for network troubleshooting.
[EXTERNAL_DOWNLOADS]: The instructions prompt the user to install or update the gmgn-cli tool using npm install -g gmgn-cli if version issues are detected.

gmgn-contract-dd