judge — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited judge (Agent Skill) and scored it 82/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 2 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
The text {match} is the classic direct prompt-injection phrasing. Placed in a skill body that the agent reads as trusted instructions, it tries to make the agent abandon its prior rules and follow whatever comes next — a full system-prompt override.
ignore/disregard/forget … previous instructions sentence.The text {match} is the classic direct prompt-injection phrasing. Placed in a skill body that the agent reads as trusted instructions, it tries to make the agent abandon its prior rules and follow whatever comes next — a full system-prompt override.
ignore/disregard/forget … previous instructions sentence.Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Multi-model council for comprehensive analysis. Adapts to whichever API keys you have configured.
/judge [your question or code]When user invokes /judge [query]:
Before calling any tool, check which tachibot-mcp tools are available in this session by looking at the loaded MCP tools. The pipeline adapts:
grok_search + grok_reasonperplexity_ask for searchopenai_reason as first judgegemini_analyze_text as final judgeqwen_coder for code analysiskimi_thinking for step-by-stepdeepseek_reason (frontier reasoning) + glm_reason (agentic) for lab diversityMinimum viable council: any 2 models from different providers. If only 1 model available: run it directly (no council needed).
Call ALL available search tools in parallel:
mcp__tachibot-mcp__grok_search (preferred - real-time)mcp__tachibot-mcp__perplexity_ask (citations, academic)mcp__tachibot-mcp__openai_search (GPT grounding)mcp__tachibot-mcp__gemini_search (Google grounding)Skip any that aren't available. If NO search tools available, proceed without search grounding and note it.
Call ALL available analysis tools in parallel:
Reasoning (pick all available):
mcp__tachibot-mcp__grok_reason - first principles analysismcp__tachibot-mcp__kimi_thinking - step-by-step, edge casesmcp__tachibot-mcp__deepseek_reason - frontier open-weight reasoning/mathmcp__tachibot-mcp__glm_reason - agentic reasoning (SWE-Bench Pro leader)Code (if query involves code):
mcp__tachibot-mcp__qwen_coder - code review and bugsmcp__tachibot-mcp__qwen_algo - algorithm optimizationUse the FIRST available from this priority list:
mcp__tachibot-mcp__openai_reason (GPQA 87.7%)mcp__tachibot-mcp__gemini_analyze_text (1M context)mcp__tachibot-mcp__grok_reason (strong reasoning)Prompt: "Extract key insights from each expert, identify consensus and conflicts. Then apply PRE-MORTEM: assume the recommended approach FAILED — what were the top 3 causes? Are those causes addressed in the analysis? Provide preliminary verdict with failure risks."
Use a DIFFERENT model from Step 3. Priority:
mcp__tachibot-mcp__gemini_analyze_text (if not used in step 3)mcp__tachibot-mcp__openai_reason (if not used in step 3)Prompt: Review first judgment, synthesize best elements from ALL analyses, resolve conflicts, provide final answer with confidence level.
Format:
/judge how to implement rate limiting/judge analyze this code: [paste]/judge microservices vs monolith for 10M users~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.