call-recording-analyzer
Pass
Audited by Gen Agent Trust Hub on Jul 4, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests call transcripts and AI-generated summaries from Clari as defined in Stage 1a of the workflow. Because transcripts contain dialogue from external parties, they represent an untrusted input surface susceptible to indirect prompt injection. Malicious instructions embedded in the transcript could potentially manipulate the MEDDIC scoring results or influence the content of Gmail drafts. The instructions lack boundary markers or explicit sanitization requirements to isolate this external data.\n
- Ingestion points:
clari_get_call_summary(SKILL.md Stage 1a)\n - Boundary markers: Absent\n
- Capability inventory: HubSpot search, Clari metadata retrieval, Gmail draft creation (config.json)\n
- Sanitization: Absent\n- [COMMAND_EXECUTION]: The skill is instructed to perform an automated file write operation to a persistent local path at
~/.claude/skill-analytics/last-outcome-call-recording-analyzer.json. While used for tracking execution metrics, this involves recurring writes to a hidden directory in the user's home folder.
Audit Metadata