The Agent Skills Directory

[SAFE]: The skill is designed for technical documentation and project management, following best practices for context preservation without introducing malicious behavior.
[COMMAND_EXECUTION]: The instructions suggest running standard local development commands (git log, git status) and a repository-specific utility script (fdk/tools/medic.py) to collect status information. These are legitimate local operations used to verify state before generating documentation.
[INDIRECT_PROMPT_INJECTION]: The skill inherently processes session context to generate handover files. While this creates a theoretical surface for passing instructions to future sessions, the skill's strict template (requiring measurements, specific blocker identification, and evidence) forces the agent to filter and structure information rather than verbatim copying, which mitigates the risk of adversarial instruction carry-over.
[DATA_EXPOSURE]: The skill writes information to local files in the llmwiki/wiki/ directory. It does not perform network operations or attempt to access sensitive system files like SSH keys or environment secrets.

orca-handover