The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The documentation in references/cli-actorization.md includes a Dockerfile example that installs the ubi tool by piping a script from a remote GitHub repository directly to the shell (curl ... | sh). This execution method bypasses integrity checks and executes unverified code from https://raw.githubusercontent.com/houseabsolute/ubi/master/bootstrap/bootstrap-ubi.sh. Notably, this contradicts the security warning in the main SKILL.md file which advises against this exact installation pattern.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of various official platform tools and libraries, including the Apify CLI, the Apify SDK for Node.js and Python, and supporting utilities like jq. These resources are fetched from standard package registries (NPM/PyPI).
[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple system-level commands to initialize, test, and deploy code (apify init, apify run, apify push). It also guides the creation of shell wrapper scripts (start.sh) that perform environment setup and binary execution.
[PROMPT_INJECTION]: The skill identifies a significant attack surface for indirect prompt injection because the Actors it creates are designed to ingest untrusted data from the web. It provides mandatory safety guidelines for sanitizing crawled content and isolating credentials from data pipelines to mitigate this risk.
Ingestion points: Project files and external web content crawled by the generated Actors.
Boundary markers: The instructions recommend using parameterized APIs and explicit sanitization logic.
Capability inventory: Includes file system access, network operations, and subprocess execution via the Apify CLI and SDK.
Sanitization: Explicitly instructs users to sanitize HTML, URLs, and text before use in sensitive operations.

apify-actorization