call-analyzer
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from emails and transcripts which can contain hidden instructions.
- Ingestion points: External content enters the agent's context through pasted conversation content (SKILL.md) and tool-retrieved email threads or call transcripts (task-method.md).
- Boundary markers: The skill lacks explicit physical delimiters for untrusted content, though it instructs the agent to parse source content before asking for facts and to distinguish speaker roles (task-method.md).
- Capability inventory: The skill has tool access to read CRM opportunity/account data, call recordings, and Octave library entities; it also produces HTML output and visual artifacts (SKILL.md, task-method.md).
- Sanitization: The agent is instructed to use an output-readiness protocol and a review protocol for artifact delivery to ensure audience suitability and evidence validation (SKILL.md).
- [COMMAND_EXECUTION]: The skill uses a tool-specific invocation pattern /octave:call-analyzer and instructs the agent to resolve tool schemas and hydrate entities from internal databases (SKILL.md, task-method.md).
Audit Metadata