The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions include patterns that download and execute remote scripts directly in a shell environment, which can allow for arbitrary code execution if the source is compromised.
Evidence: curl -fsSL https://agentsview.io/install.sh | bash in SKILL.md.
Evidence: powershell -ExecutionPolicy ByPass -c "irm https://agentsview.io/install.ps1 | iex" in SKILL.md.
[PERSISTENCE]: The skill encourages users to modify their shell startup files and create background system services, which allows code to persist and run automatically across sessions.
Evidence: Instructions to add an analytics function to shell profiles (~/.bashrc or ~/.zshrc) to monitor costs in every new terminal prompt.
Evidence: Example configuration for a systemd user service (~/.config/systemd/user/agentsview.service) to run the tool as a background daemon.
[INDIRECT_PROMPT_INJECTION]: The skill processes session data from various coding agents and metrics from GitHub, creating a surface where malicious instructions embedded in those external sources could be interpreted by the agent.
Ingestion points: Agent session directories such as ~/.claude/projects and ~/.cursor/sessions specified in SKILL.md, as well as GitHub PR outcomes.
Boundary markers: None present; no instructions are provided to the agent to ignore or delimit potentially untrusted content within ingested sessions.
Capability inventory: The skill can execute various CLI commands (agentsview sync, stats, serve), access the local filesystem, and manage SQLite/PostgreSQL databases.
Sanitization: No sanitization or validation steps are documented for the imported session data.
[PRIVILEGE_ESCALATION]: The skill utilizes commands designed to bypass local execution security policies.
Evidence: Use of -ExecutionPolicy ByPass in the Windows PowerShell installation command.
[COMMAND_EXECUTION]: The tool relies on a wide array of local CLI commands and background server processes, which increases the system's attack surface if the binary is compromised.
[EXTERNAL_DOWNLOADS]: The skill fetches installation scripts from a non-trusted external domain (agentsview.io). Note that it also uses official registries for Docker images (ghcr.io) and Homebrew, which are standard services.

agentsview-local-agent-analytics