weekly-intel-digest

Pass

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from shared environments.
  • Ingestion points: Untrusted data enters the agent context via message history from five shared Slack channels (#feed-swyft-meetings, #feed-swyft-customer-quotes, #feed-competitor-feedback, #feed-swyft-churn-risks, and #feed-twitter) and via DM responses from internal bots (tiger-analytics and eon).
  • Boundary markers: The skill lacks explicit boundary markers or delimiters to isolate untrusted data, increasing the risk that malicious instructions embedded in Slack messages could influence the agent's behavior.
  • Capability inventory: The skill has the capability to search all public and private Slack channels (slack_search_public_and_private), access the current user's profile (slack_read_user_profile), and send direct messages (slack_send_message).
  • Sanitization: There is no evidence of content sanitization or validation performed on the ingested Slack data before it is processed and summarized.
Audit Metadata
Risk Level
SAFE
Analyzed
May 5, 2026, 07:20 AM