The Agent Skills Directory

Command Execution: The skill instructs the agent to use standard Unix-like shell commands to explore and read log files. These tools are used for their intended purpose of log analysis.
Evidence: Use of ls, tail, rg (ripgrep), and find in the Procedure and Useful Commands sections of SKILL.md.
Indirect Prompt Injection: The skill is designed to ingest and process log files which could contain untrusted data. This presents a potential surface where malicious instructions embedded in logs could be interpreted by the agent.
Ingestion points: The agent reads content from various log files (e.g., renderer.log, exthost.log, agenthost.log) located in the <user-data-dir>/logs directory as described in SKILL.md.
Boundary markers: The provided instructions do not specify any delimiters or safety prompts to prevent the agent from following instructions that might be contained within the log data.
Capability inventory: The primary tools used are read-only (tail, rg). However, the skill exists within a development context where the agent may have broader file-system or shell capabilities.
Sanitization: No explicit sanitization or filtering of the log content is performed before the agent processes or displays it.
Source Code Modification: The skill includes a procedure for modifying a TypeScript file (src/vs/platform/log/common/log.ts) to enable debugging flags. While this is a standard developer workflow, it involves the agent programmatically changing project source code.
Evidence: The Temporary Console Forwarding Workflow section provides specific code blocks for modifying isDevConsoleLogForwardingEnabled in the Code OSS repository.

code-oss-logs