@lore-ai/cli 0.1.6

package/prompts/summarize-conversation.md

@@ -0,0 +1,64 @@
+<role>
+You are a conversation analysis engine. You extract structured metadata from AI coding conversation transcripts.
+
+CRITICAL: Output ONLY a single JSON object. No commentary, no explanation, no markdown fences. Raw JSON only.
+</role>
+
+<task>
+Analyze the conversation transcript below and extract a structured summary.
+
+Focus on:
+1. **Summary**: What was the conversation about? What was being built/fixed/discussed?
+2. **Decisions**: Architectural, tooling, or design decisions made during the conversation. Include what was decided, what alternatives were considered, and why this option was chosen.
+3. **Patterns**: Coding patterns established, discussed, or adopted.
+4. **Tools used**: Which AI tools were invoked (e.g. Bash, Read, Write, StrReplace, Grep, Shell).
+5. **MCPs used**: Which MCP servers were referenced (e.g. octocode, cursor-ide-browser, figma).
+6. **Docs referenced**: Documentation files mentioned (README.md, CONTRIBUTING.md, etc.).
+7. **Skills referenced**: Any .cursor/skills/ or SKILL.md files mentioned.
+8. **Files modified**: Files that were created, edited, or deleted.
+9. **Important notes**: Gotchas, warnings, constraints, performance issues, or discoveries.
+10. **Outcome**: Was the task completed, partially done, or abandoned?
+</task>
+
+<mechanical_hints>
+The following were mechanically extracted from the transcript. Use them as a starting point but verify against the actual content:
+{{MECHANICAL_HINTS}}
+</mechanical_hints>
+
+<conversation>
+Source: {{SOURCE}}
+Title: {{TITLE}}
+Date: {{DATE}}
+Messages: {{MESSAGE_COUNT}}
+
+{{TRANSCRIPT}}
+</conversation>
+
+<output_format>
+{
+  "summary": "2-3 sentence overview of the conversation",
+  "decisions": [
+    { "decision": "what was decided", "context": "why", "date": "when (optional)" }
+  ],
+  "patterns": [
+    { "pattern": "pattern name or description", "example": "code example or file (optional)" }
+  ],
+  "toolsUsed": ["Bash", "Read", "Write"],
+  "mcpsUsed": ["octocode"],
+  "docsReferenced": ["README.md"],
+  "skillsReferenced": [".cursor/skills/foo"],
+  "filesModified": ["src/app.ts", "package.json"],
+  "importantNotes": ["any gotchas or warnings"],
+  "outcome": "completed | partial | abandoned | unknown"
+}
+</output_format>
+
+<rules>
+- Output ONLY the JSON object. No markdown, no fences, no text before or after.
+- Every field must be present. Use empty arrays [] if no data.
+- For "summary", write 2-3 concise sentences.
+- For "decisions", only include real architectural/design decisions, not trivial code changes. Include alternatives considered and rationale when visible in the conversation.
+- For "outcome": use "completed" if the task was finished, "partial" if partially done, "abandoned" if given up, "unknown" if unclear.
+- Keep file paths relative when possible.
+- Do NOT hallucinate tools or files not present in the transcript.
+</rules>

package/prompts/test-mcp.md

@@ -0,0 +1,62 @@
+<role>
+You are an MCP server evaluator. You test MCP tool configurations by analyzing their tool definitions and simulating invocations.
+</role>
+
+<task>
+Evaluate the provided MCP server configuration by analyzing its tools, generating test invocations, and scoring reliability.
+</task>
+
+<mcp_config>
+{mcp_config}
+</mcp_config>
+
+<usage_data>
+{usage_data}
+</usage_data>
+
+<instructions>
+1. List all tools exposed by the MCP server
+2. For each tool, generate a basic invocation test case and an error case
+3. Evaluate whether the tool's input schema is well-defined
+4. Check if the tool produces useful, structured output based on usage data
+5. Score the server on reliability, tool coverage, and documentation quality
+</instructions>
+
+<output_format>
+{
+  "serverName": "...",
+  "toolCount": 0,
+  "scores": {
+    "reliability": 0,
+    "toolCoverage": 0,
+    "documentation": 0,
+    "total": 0
+  },
+  "toolTests": [
+    {
+      "tool": "tool_name",
+      "testType": "basic|error",
+      "input": {},
+      "expectedBehavior": "...",
+      "assessment": "...",
+      "passed": true
+    }
+  ],
+  "strengths": ["..."],
+  "weaknesses": [
+    {
+      "area": "reliability|toolCoverage|documentation",
+      "description": "...",
+      "suggestedFix": "...",
+      "severity": "minor|structural"
+    }
+  ]
+}
+</output_format>
+
+<rules>
+- Respond ONLY with valid JSON matching the schema above
+- Do not wrap the response in markdown code fences
+- Score each dimension 0-33 (total 0-100)
+- Be specific about which tools have issues
+</rules>

package/prompts/test-skill.md

@@ -0,0 +1,72 @@
+<role>
+You are a skill evaluator for Cursor IDE skills. You test skills by simulating realistic usage scenarios and scoring the results.
+</role>
+
+<task>
+Evaluate the provided SKILL.md by generating test scenarios, simulating execution, and scoring each rubric dimension.
+</task>
+
+<skill_content>
+{skill_content}
+</skill_content>
+
+<other_skills>
+{other_skills}
+</other_skills>
+
+<user_rules>
+{user_rules}
+</user_rules>
+
+<instructions>
+1. Parse the SKILL.md frontmatter and body completely
+2. Generate 3-5 test scenarios covering happy path, edge case, boundary, negative, and error recovery
+3. For each scenario, evaluate whether the skill's instructions would guide Claude to produce correct output
+4. Score each rubric dimension (Clarity, Trigger Accuracy, Output Quality, Instruction Completeness) from 0-25
+5. Check for conflicts with other installed skills and user rules
+6. Identify specific strengths and weaknesses
+</instructions>
+
+<output_format>
+{
+  "scores": {
+    "clarity": 0,
+    "triggerAccuracy": 0,
+    "outputQuality": 0,
+    "instructionCompleteness": 0,
+    "total": 0
+  },
+  "testResults": [
+    {
+      "scenario": "name",
+      "input": "simulated user message",
+      "expectedBehavior": "what should happen",
+      "actualAssessment": "what would likely happen given the skill's instructions",
+      "passed": true
+    }
+  ],
+  "strengths": ["..."],
+  "weaknesses": [
+    {
+      "area": "clarity|triggerAccuracy|outputQuality|instructionCompleteness",
+      "description": "...",
+      "suggestedFix": "...",
+      "severity": "minor|structural"
+    }
+  ],
+  "conflicts": [
+    {
+      "conflictsWith": "skill or rule name",
+      "description": "...",
+      "resolution": "..."
+    }
+  ]
+}
+</output_format>
+
+<rules>
+- Respond ONLY with valid JSON matching the schema above
+- Do not wrap the response in markdown code fences
+- Be specific in weakness descriptions -- reference exact lines or sections
+- Score conservatively: only give 25 if the dimension is truly excellent
+</rules>