The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/scan_deepseek_sessions.py and scripts/profile_store.py) to perform its primary function of scanning session logs and managing local configuration context. This is expected behavior for its diagnostic and session-recovery purpose.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an indirect injection surface as it processes local session logs which may contain untrusted data from previous sessions. Although this surface exists, the risk is negligible due to the diagnostic constraints.\n
Ingestion points: The agent reads local rollout files (.jsonl) from the Codex session directory via scripts/scan_deepseek_sessions.py.\n
Boundary markers: The instructions focus the agent strictly on technical diagnostic markers (tool call IDs, status codes, and timestamps) rather than interpreting natural language prompts, which significantly reduces the risk of malicious instruction obedience.\n
Capability inventory: The skill can execute local scripts, read session logs, and update a local profile JSON file.\n
Sanitization: While no content-level sanitization is performed on the raw log text, the scanner script pre-filters for specific technical JSON keys and line markers.\n- [EXTERNAL_DOWNLOADS]: The validation utility scripts/validate_skill.py identifies a requirement for the standard PyYAML library and suggests its installation through official package managers, which is a common and safe development practice.

lov-fix-deepseek-tool-call-error