influence-and-negotiation
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process vast amounts of untrusted data from external sources, creating a risk that malicious instructions embedded in that data could influence the agent's behavior.
- Ingestion points: Untrusted data enters the agent context through the intake of external emails, Slack messages, and meeting transcripts in Step 1 of the context intake process, and through automated profiling of stakeholders using web search and LinkedIn in Step 2 of
references/context-intake.md. - Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing raw material from external counterparties.
- Capability inventory: The skill is granted permissions for
Read,Write,Edit,Agent,WebFetch, andWebSearchacross its instruction files. - Sanitization: Absent. There are no instructions provided for escaping, validating, or filtering the content retrieved from external web sources or messages before it is processed by the agent or its sub-agents.
- Ingestion points: Untrusted data enters the agent context through the intake of external emails, Slack messages, and meeting transcripts in Step 1 of the context intake process, and through automated profiling of stakeholders using web search and LinkedIn in Step 2 of
- [EXTERNAL_DOWNLOADS]: The skill frequently uses web search and fetch tools to gather intelligence on companies, decision-makers, and market benchmarks.
- Evidence:
SKILL.mdincludesWebSearchandWebFetchin itsallowed-tools.references/context-intake.mdStep 2 mandates deep research using these tools to profile stakeholders across LinkedIn, news archives, and regulatory databases.
- Evidence:
Audit Metadata