The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands using placeholders for user-supplied data, such as {curator name}, {curator slug}, and {test file or test name}. For example, rg -n "{curator name}|{curator slug}" and poetry run pytest {test file or test name} -v. If an attacker provides a curator name or test name containing shell metacharacters (e.g., ;, |, `), it could lead to arbitrary command execution in the user's environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted metadata (Curator name, website, social links) and incorporates it into the agent's logic, including code generation for eth_defi/vault/curator.py and YAML configuration files.
Ingestion points: User-provided inputs for curator metadata (Step 1-3).
Boundary markers: None present; user input is directly interpolated into instructions and shell commands.
Capability inventory: Subprocess execution via poetry, rg, and pytest; file writing/editing capability for project source code and configuration files.
Sanitization: No validation or escaping of the user-provided inputs is specified before they are used in shell commands or code edits.

add-curator