platform-capability-search

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes pre-defined internal scripts from the Salesforce plugin root directory (${CLAUDE_PLUGIN_ROOT}/scripts/sf-context) to perform project discovery, journey tracking, and feature detection.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external metadata from plugin registries which could contain malicious instructions.
  • Ingestion points: sf-context plugin-match command output containing third-party plugin descriptions and summaries in SKILL.md.
  • Boundary markers: The skill explicitly instructs the agent to treat this data as "untrusted metadata" and forbids following any instructions found within it.
  • Capability inventory: Accesses the Bash tool to run local scripts (sf-context).
  • Sanitization: The instructions specifically mandate that only fixed commands in the skill are executable and that registry text must never be followed.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 12:41 PM
Security Audit — agent-trust-hub — platform-capability-search