The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/runtime_bootstrap.py uses os.execve to restart the current script within a dedicated Python virtual environment located at ~/.claude/.fetching-salesforce-docs-runtime. While os.execve is a powerful primitive for process replacement, in this context, it is used for local environment management to ensure dependencies like Playwright are available.
[EXTERNAL_DOWNLOADS]: The skill documentation indicates that the installer automatically downloads and installs the Playwright Chromium browser into an isolated runtime directory. This is a standard requirement for the playwright library used by the extraction scripts.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it retrieves and processes content from the public web.
Ingestion points: Content is fetched from Salesforce-owned domains including developer.salesforce.com, help.salesforce.com, architect.salesforce.com, and lightningdesignsystem.com via the extract_salesforce_doc.py and extract_help_salesforce.py scripts.
Boundary markers: Not explicitly defined in the provided scripts or prompts.
Capability inventory: The skill has the capability to perform network requests (via Playwright) and execute system-level commands (via os.execve for bootstrapping).
Sanitization: The scripts implement text normalization and cleanup (e.g., cleanup_help_text, normalize_text) to remove navigation noise and shell elements, but they do not specifically filter for malicious instructions embedded in the documentation body.

fetching-salesforce-docs