The Agent Skills Directory

[DYNAMIC_EXECUTION]: The runtime_bootstrap.py script utilizes os.execve() to replace the current process with a Python interpreter from the skill's managed virtual environment located at ~/.claude/.fetching-salesforce-docs-runtime. Additionally, extraction scripts use page.evaluate() to execute JavaScript within the browser context to navigate complex Salesforce documentation components.
[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of verified Python packages (playwright, playwright-stealth) and browser binaries into its isolated local runtime directory during the setup phase.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Salesforce URLs (ingestion points: extract_help_salesforce.py and extract_salesforce_doc.py). It implements boundary markers via 'Rejection Rules' in SKILL.md and uses heuristic cleaning (cleanup_help_text) to sanitize navigation and shell noise, reducing the risk from untrusted documentation content.
[SAFE]: The skill scripts perform strict host validation to ensure that network requests are only directed to official Salesforce-owned domains, preventing the skill from being misused for arbitrary web scraping.

platform-docs-get