The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @monid-ai/cli package from the NPM registry and instructs the agent to download an updated version of SKILL.md from https://monid.ai/SKILL.md to maintain synchronization with the CLI tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from a wide variety of external sources (e.g., social media, search results, company data) which could contain adversarial content.
Ingestion points: Data is retrieved via monid run commands and often saved to local files (e.g., results.json) for the agent to analyze.
Boundary markers: There are no specific delimiters or instructions provided to separate system prompts from the untrusted data being processed.
Capability inventory: The skill allows for command execution via the CLI, local file writing, and extensive network interactions with various data providers.
Sanitization: No explicit validation or filtering mechanisms are defined for handling the content returned by external APIs.
[DATA_EXFILTRATION]: The skill provides a flow for uploading local files to a vendor-controlled remote file system (sfs.monid.ai) using signed URLs to enable access for third-party tools.
[COMMAND_EXECUTION]: Workflows include example shell commands that use command substitution (e.g., $(wc -c < file)) to obtain file metadata during processing.
[CREDENTIALS_UNSAFE]: The agent is instructed to manage API keys by prompting the user to provide them and then storing them using the monid keys add command.

pay-per-call-apis