platform-capability-search

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external registry metadata and user-provided task descriptions, which are potential ingestion points for malicious instructions. It mitigates this risk by explicitly instructing the agent to treat all registry descriptions, examples, and summaries as untrusted metadata and to never follow that text as instructions or execute commands found within it. Boundary markers are present in the form of instructions to only execute the fixed commands defined in the skill.
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute project-internal scripts located at ${CLAUDE_PLUGIN_ROOT}/scripts/sf-context. These commands are used for discovery, feature detection, and journey management. The skill implements a security-conscious workflow for sensitive operations like history resets, requiring a dry-run to generate a nonce which the user must then explicitly confirm before the operation is executed.
  • [DATA_EXPOSURE]: The skill accesses Salesforce project and org metadata to provide journey hints and feature detection. It includes specific instructions to ensure that raw invalid content, hashes, and sensitive paths are never shown in the output, and that CLI responses are sanitized before being presented to the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 04:51 PM