The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides an installation command curl -fsS https://cli.dflow.net | sh. This is a high-risk execution pattern that downloads a script from a remote URL and pipes it directly into a shell interpreter. This allows the remote server to execute arbitrary code on the user's system without prior verification or integrity checks.
[EXTERNAL_DOWNLOADS]: The skill downloads a documentation index from https://pond.dflow.net/llms.txt and saves it to a local path (skills/dflow-docs/llms.txt). It also references a hosted MCP server at https://pond.dflow.net/mcp. These actions create a runtime dependency on external vendor infrastructure.
[COMMAND_EXECUTION]: The skill uses several shell commands to manage its lifecycle, including curl for networking, sh for script execution, and sed for reading local documentation files. It also utilizes the dflow CLI for platform interactions.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by downloading a remote text file (llms.txt) and instructing the agent to read its contents into the current context.
Ingestion points: skills/dflow-docs/llms.txt (referenced and updated via curl and sed commands in SKILL.md)
Boundary markers: None identified; external content is read directly into context.
Capability inventory: Subprocess execution (sh), file system writes (curl -o), and network operations (curl).
Sanitization: No sanitization or validation of the downloaded content is performed before the agent processes it.

dflow-docs