Route Specialist

Classifies tasks by domain using deterministic and LLM methods, then routes to specialized prompts with tuned models and metadata for expert handling.

What to check before installing: - Clarify the env-var mismatch: the SKILL.md lists OPENCLAW_LLM_ENDPOINT under metadata but the registry says no required env vars. Confirm whether an LLM endpoint/credential is needed and how it's provided/stored. - Review the full specialist system prompts (they were truncated). Those prompts will be sent to models and can contain instructions that affect outputs or data handling. - Limit the file_context you pass to the skill. It uses file names/extensions (reasonable) but the skill has access to the file-system tool — avoid giving paths to sensitive files. - Ask the publisher to explain the model choice/comments (M2.7 vs gemma-4-26b-moe and the MBP M1 note) to confirm there are no hidden assumptions about runtime environment. - Because the skill can call the LLM autonomously, only grant it to agents/workflows you trust; if you need stricter control, disable autonomous invocation or audit calls/logs. If the publisher clarifies the env-var declaration and you inspect the full prompts, the inconsistencies appear explainable; until then treat the skill with caution.

Type: OpenClaw Skill Name: route-specialist Version: 1.0.0 The 'route-specialist' skill is a legitimate task router designed to improve AI performance by selecting domain-specific system prompts (e.g., iOS, Trading, DevOps). The logic in SKILL.md uses deterministic file-extension matching and LLM-based classification to route tasks to appropriate models and prompts. The specialist prompts themselves promote security best practices, such as forbidding the use of 'eval' or 'exec' in Python and enforcing least-privilege IAM in DevOps. No indicators of data exfiltration, malicious execution, or prompt injection were found.