The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/runtime_bootstrap.py uses os.execve to re-execute extraction tasks within an isolated virtual environment located at ~/.claude/.fetching-salesforce-docs-runtime. This is a standard and safe implementation for managing specific library dependencies like Playwright without impacting the host system's global environment.
[EXTERNAL_DOWNLOADS]: The skill is configured to fetch documentation content from official and well-known Salesforce domains, including developer.salesforce.com, help.salesforce.com, architect.salesforce.com, and lightningdesignsystem.com. These operations are consistent with the skill's stated purpose and target trusted organizational resources.
[INDIRECT_PROMPT_INJECTION]: The skill processes external HTML content from documentation pages, creating a potential surface for indirect prompt injection.
Ingestion points: Content is fetched from external Salesforce URLs via the scripts in the scripts/ directory.
Boundary markers: SKILL.md provides strict classification and acceptance rules, instructing the agent to verify specific identifiers and reject landing pages or shell content.
Capability inventory: The extraction scripts are read-only and output text to the console for the agent's context; they do not perform file writes or execute commands based on the retrieved content.
Sanitization: The Python scripts include cleanup_help_text and normalize_text functions to filter out navigation noise, banners, and shell tokens before the content is presented to the agent.

platform-docs-get