The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides installation commands that pipe remote scripts directly into system shells for execution. Examples include fetching from astral.sh and piping to 'sh' on macOS and 'iex' on Windows.
[PRIVILEGE_ESCALATION]: The Windows installation instructions utilize the 'powershell -ExecutionPolicy ByPass' flag to circumvent default security restrictions and execute the remote installer script.
[PERSISTENCE]: The skill includes a setup command ('inf setup') that is designed to overwrite its own files in the local directory ('~/.claude/skills/informe-cli/'), which constitutes self-modifying behavior and persistence across updates.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from a wide variety of untrusted external sources, including Slack, Notion, Gmail, and production databases.
Ingestion points: Processes data from internal databases via SQL queries, Slack messages, Notion pages, and local files provided via the '--attach' flag.
Boundary markers: No specific delimiters or warnings to ignore instructions within the retrieved data were identified in the instructions.
Capability inventory: Executes the 'inf' CLI tool, performs network requests to 'informe-agent.mindlogic.tools', and handles file operations such as 'inf files get'.
Sanitization: The skill does not document any sanitization or filtering processes for the external content it processes.
[EXTERNAL_DOWNLOADS]: The skill initiates downloads of tools and configuration from well-known services and vendor-specific repositories, including 'uv' from Astral, 'cloudflared' from Cloudflare, and the CLI tool from the vendor's GitHub organization.

informe-cli