Objection to Playbook Mapper
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No patterns of instruction override, jailbreak attempts, or safety filter bypasses were identified. The skill maintains a consistent focus on its stated sales-enablement purpose.
- [DATA_EXFILTRATION]: The skill uses platform capabilities to access CRM data and meeting transcripts. These operations are restricted to internal tools (
get_deal,get_company_status,search_transcripts) and do not involve unauthorized network requests to external domains. No hardcoded credentials or sensitive file path access (e.g., .ssh, .env) were found. - [REMOTE_CODE_EXECUTION]: The skill does not perform any remote script downloads, package installations, or dynamic code evaluation. It is entirely prompt-based and relies on structured output generation.
- [COMMAND_EXECUTION]: No shell commands, subprocess spawning, or administrative privilege escalations (e.g., sudo, chmod) are present in the skill's logic.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (the 'objection' text) and internal data (meeting transcripts). While there are no explicit boundary markers like XML tags, the risk is mitigated by the skill's strict output contract and lack of dangerous capabilities (such as shell access or file system modification). The analysis of input data is restricted to classification and response generation.
- [OBFUSCATION]: The content is clear and readable markdown. No Base64 encoding, zero-width characters, homoglyphs, or hidden tag characters were detected.
Audit Metadata