The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/workspace_snapshot.py performs shell execution via subprocess.run to gather git information. The calls are constructed safely using lists of arguments, mitigating command injection risks.\n- [DATA_EXFILTRATION]: No network-enabled tools or exfiltration patterns were found. The skill operates exclusively on the local filesystem and repository metadata.\n- [PROMPT_INJECTION]: Analysis of SKILL.md confirms the instructions are benign and aimed solely at providing session summaries. There are no attempts to manipulate the agent's core instructions or safety filters.

session-recap