agentforce-bot-upgrade

Pass

Audited by Gen Agent Trust Hub on Nov 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by ingesting natural language data from Einstein Bot metadata and mapping it into Agentforce agent instructions during the migration process.
  • Ingestion points: Reads bot metadata files such as .bot-meta.xml and .botVersion-meta.xml, as well as associated Apex and Flow source code, as specified in references/input-contract.md and references/extraction-blueprint.md.
  • Boundary markers: The mapping logic in references/mapping-rules.md and references/handoff-output-format.md does not describe the use of explicit delimiters or instructions to ignore potential commands embedded within the ingested metadata.
  • Capability inventory: The skill utilizes the Salesforce CLI (sf) for data queries, metadata deployment, and agent publishing, providing a significant capability set if instructions are successfully manipulated.
  • Sanitization: No explicit sanitization or filtering of natural language content retrieved from the source bot is defined in the extraction or mapping workflows.- [EXTERNAL_DOWNLOADS]: Downloads and installs the AgentScript SDK from official Salesforce-maintained repositories during initialization.
  • Evidence: The references/_vendored/agentscript-toolchain.md file specifies downloading the @sf-agentscript/agentforce package from the npm registry or cloning from github.com/salesforce/agentscript.git. These are recognized as trusted vendor sources, representing standard functionality for this developer tool.- [COMMAND_EXECUTION]: Executes various shell commands and scripts to automate the bot migration workflow and interact with Salesforce orgs.
  • Evidence: The skill instructions in SKILL.md and multiple reference files utilize node, npm, and the Salesforce CLI (sf) for project generation, metadata retrieval, and agent deployment. These operations are consistent with the skill's primary purpose and the requirements of the Salesforce development ecosystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 8, 2026, 11:45 AM