The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides installation instructions that pipe remote scripts directly into shell interpreters (curl -fsSL https://agentsview.io/install.sh | bash and powershell -ExecutionPolicy ByPass -c "irm https://agentsview.io/install.ps1 | iex"). This allows for the execution of unverified arbitrary code from a non-trusted third-party domain.\n- [COMMAND_EXECUTION]: The Windows installation instructions include the use of -ExecutionPolicy ByPass, which overrides security settings designed to prevent the execution of untrusted scripts. Additionally, the skill provides instructions for setting up a persistent background service via systemd (agentsview.service).\n- [EXTERNAL_DOWNLOADS]: The skill directs users to download software, Docker images, and source code from external, non-trusted sources such as agentsview.io and the kenn-io GitHub/GHCR repositories.\n- [CREDENTIALS_UNSAFE]: Configuration examples demonstrate hardcoding sensitive information like PostgreSQL connection strings with passwords and API tokens into environment variables (AGENTSVIEW_PG_URL, AGENTSVIEW_DUCKDB_TOKEN). Following these examples can lead to credential exposure in shell histories or local files.\n- [PROMPT_INJECTION]: The skill processes untrusted data from AI agent session histories (e.g., .claude/projects, .cursor/sessions) which may contain malicious instructions.\n
Ingestion points: Syncs and analyzes agent session directories.\n
Boundary markers: No specific delimiters or warnings to ignore embedded content are provided.\n
Capability inventory: The tool can perform file system operations (sync), network communications (PostgreSQL/DuckDB), and serve a local web UI.\n
Sanitization: No evidence of sanitization or filtering of the ingested session data is present in the documentation.

agentsview-local-agent-analytics