The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a series of subcommands via ${CLAUDE_PLUGIN_ROOT}/scripts/sf-context. These commands are scoped to discovery, journey tracking, and organizational feature detection within the Salesforce ecosystem.
[EXTERNAL_DOWNLOADS]: The skill manages the installation of new capabilities using npx skills@1.5.20 to pull from the forcedotcom/sf-skills repository. This process is protected by multiple security layers, including version pinning and the requirement for a specific, non-reconstructible command format.
[REMOTE_CODE_EXECUTION]: The skill provides a guarded mechanism to add new tools. This capability is secured by a validation step that checks the skill name against a strict regular expression (^[a-z0-9]+(-[a-z0-9]+)*$) and ensures the metadata matches a predefined 'publicAvailable' status before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes capability catalogs which may contain untrusted data. It implements a mandatory evidence chain for safety:
Ingestion points: Data enters the context via the sf-context discovery commands.
Boundary markers: The instructions explicitly designate all catalog descriptions and summaries as 'untrusted metadata' and forbid the agent from following instructions found within them.
Capability inventory: The skill has access to shell execution via the Bash tool and the npx installer.
Sanitization: All skill names are validated via regex, and a confirmation nonce mechanism is used for high-impact actions like resetting the project journey history.

platform-capability-search