The Agent Skills Directory

[COMMAND_EXECUTION]: The file scripts/runtime_bootstrap.py utilizes os.execve to transition script execution to a specific Python interpreter within a dedicated virtual environment directory (~/.claude/.fetching-salesforce-docs-runtime/venv). This process-replacement technique is used for environment management and to ensure the skill runs with its required dependencies.
[COMMAND_EXECUTION]: The extraction scripts (scripts/extract_help_salesforce.py and scripts/extract_salesforce_doc.py) use the Playwright library to execute a Chromium browser instance to scrape content. This involves executing browser binaries locally.
[EXTERNAL_DOWNLOADS]: The skill requires the playwright and playwright-stealth Python packages. Installation of these dependencies involves fetching code from official package registries and browser binaries from trusted provider servers.
[PROMPT_INJECTION]: The skill processes external web content from official Salesforce domains, creating an indirect prompt injection surface.
Ingestion points: Data enters the agent's context through scripts/extract_help_salesforce.py and scripts/extract_salesforce_doc.py which scrape web pages.
Boundary markers: No explicit markers are used in the scraped content, although SKILL.md provides interpretation rules.
Capability inventory: The skill can execute subprocesses via Playwright and replace its own process via os.execve.
Sanitization: The scripts normalize text and remove layout-related noise but do not implement specific sanitization to prevent the agent from following instructions embedded in the scraped HTML.

platform-docs-get