The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill implements an automated credential management system in references/cookie-vault.md and SKILL.md. It explicitly instructs the AI agent to harvest session cookies and authentication tokens from web interactions and persist them in a plaintext markdown file for reuse in future scraping tasks.
[COMMAND_EXECUTION]: The operational workflow involves the agent using Bash to execute Python scripts. These scripts are generated at runtime by injecting user-provided data (URLs, CSS selectors, login data) into pre-defined templates such as templates/basic_fetch.py and templates/session_login.py through string replacement.
[EXTERNAL_DOWNLOADS]: The maintenance and installation instructions in references/maintenance.md require the execution of commands that download the scrapling package from PyPI and browser binaries (Camoufox and Playwright) using the scrapling install command from remote repositories.
[REMOTE_CODE_EXECUTION]: The core functionality of the skill relies on the dynamic generation and execution of Python code on the host system to perform scraping operations, which allows for the execution of arbitrary code logic based on the templates provided.
[PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it processes untrusted data from external websites.
Ingestion points: External website content is ingested via Fetcher and StealthyFetcher modules in the templates/ directory.
Boundary markers: None; the scraped data is directly handled by the agent without delimiters or instructions to ignore embedded commands.
Capability inventory: The agent has access to shell execution (Bash(python*)), network requests, and the ability to read and write to the local file system (including the cookie vault).
Sanitization: No sanitization or validation of the ingested HTML or text content is performed before it is processed by the agent.

scrapling