The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill reads and searches historical log files, creating a surface for indirect prompt injection if those logs contain malicious instructions from previous agent interactions.
Ingestion points: SESSION.md and files within sessions/ directories via read_file.
Boundary markers: Uses Markdown structure and HTML <details> tags for logical separation, though no explicit 'ignore' instructions are present.
Capability inventory: File system access (read_file, write_file); no network or subprocess execution detected.
Sanitization: No explicit sanitization or filtering of logged content is defined in the provided logic.- [Data Exposure & Exfiltration] (SAFE): While the skill requires file-writing permissions, it does not target sensitive system paths (e.g., SSH keys or environment files) and contains no network tools for exfiltration.

session-log