The Agent Skills Directory

[SAFE]: The skill uses the context: fork parameter to copy the session into an isolated environment, preventing the primary agent context from being contaminated by large-scale codebase analysis.
[SAFE]: Instructions restrict the agent to read-only tools such as Glob, Grep, and Read for structural analysis and pattern discovery, ensuring no files are modified.
[SAFE]: The skill documents an Indirect Prompt Injection surface where codebase files are ingested. Ingestion points: local files accessed via Read, Grep, and Glob. Boundary markers: absent. Capability inventory: restricted to read-only tools within the isolated Explore agent. Sanitization: none mentioned. The risk is minimized by the isolated execution environment and lack of write capabilities.
[SAFE]: No network exfiltration, hardcoded credentials, or unauthorized command execution patterns were detected in the skill's instructions or strategy guides.

research-deep