analyze-sessions
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Access to Sensitive Session Logs: The skill reads local Copilot CLI session databases and transcript files stored in the user's home directory (~/.copilot/...). This access is required for the skill to perform its intended function of analyzing agent performance and failure modes.
- External Tool Acquisition: The skill may download and execute external utilities, including dotnet-replay (via dnx) and @microsoft/vally-cli (via npx). These tools are sourced from well-known package registries and are used for log normalization and regression testing within the MAUI development workflow.
- Command Execution: The skill executes local shell commands using PowerShell (pwsh) and sqlite3 to query session metadata and process logs. These operations are performed locally and do not involve remote command execution.
- Data Processing Surface (Indirect Prompt Injection): The skill processes session transcripts, which represents an indirect prompt injection surface. The instructions include explicit boundary markers labeling transcript-derived content as untrusted evidence and direct the agent to disregard any instructions contained within those logs.
- Privacy and Redaction Measures: The included PowerShell script implements a Protect-Text function that uses regular expressions to redact sensitive information—such as authentication tokens, email addresses, and private keys—from the session analysis report before it is presented to the agent.
Audit Metadata