The Agent Skills Directory

[SAFE]: The skill uses dynamic context injection via the !command syntax to pre-compute project metadata, including the current git branch, session ID, recent commits, and working tree status. These commands are benign utility operations used solely for providing situational awareness and do not involve unauthorized file access or network exfiltration.\n- [SAFE]: The skill ingests data from local filesystem sources (handoffs and ledgers), which represents a potential surface for indirect prompt injection. However, since the skill is strictly read-only and lacks capabilities for file modification, arbitrary code execution, or network exfiltration, this surface is considered safe in this context.\n
Ingestion points: Data is read from files in <memory_dir>/handoffs/, <memory_dir>/<slug>/, and <memory_dir>/running-retros/.\n
Boundary markers: No explicit delimiters are defined in the instructions for the ingested file content.\n
Capability inventory: The skill is limited to read-only operations and has no access to dangerous tools or external network endpoints.\n
Sanitization: No explicit sanitization or filtering of the ingested file content is performed.

orient