torbjorn/my-pal-mcp-server
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Generic naming to work with Gemini CLI / Claude Code

Browse Source
This commit is contained in:
Fahad
2025-06-27 23:41:20 +04:00
parent 75dc724ecd
commit bc447d4bcd
31 changed files with 139 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
simulator_tests/test_consensus_three_models.py
View File

@@ -124,12 +124,12 @@ class TestConsensusThreeModels(BaseSimulatorTest):
self.logger.info(f"Model name in metadata: {metadata.get('model_name')}")
# Verify we have analysis from Claude
claude_analysis = consensus_data.get("claude_analysis")
if not claude_analysis:
agent_analysis = consensus_data.get("agent_analysis")
if not agent_analysis:
self.logger.error("Missing Claude's analysis in step 1")
return False
analysis_text = claude_analysis.get("initial_analysis", "")
analysis_text = agent_analysis.get("initial_analysis", "")
self.logger.info(f"Claude analysis length: {len(analysis_text)} characters")
self.logger.info("✓ Three-model consensus tool test completed successfully")

4
simulator_tests/test_consensus_workflow_accurate.py
View File

@@ -87,8 +87,8 @@ class TestConsensusWorkflowAccurate(ConversationBaseTest):
return False
# Verify Claude's analysis is included
if "claude_analysis" not in step1_data:
self.logger.error("Expected claude_analysis in step 1 response")
if "agent_analysis" not in step1_data:
self.logger.error("Expected agent_analysis in step 1 response")
return False
# Verify first model response is included

6
systemprompts/analyze_prompt.py
View File

@@ -10,7 +10,7 @@ and maintainability—not just spot routine code-review issues.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -21,7 +21,7 @@ MUST respond ONLY with this JSON format (and nothing else). Do NOT ask for the s
for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -29,7 +29,7 @@ ESCALATE TO A FULL CODEREVIEW IF REQUIRED
If, after thoroughly analysing the question and the provided code, you determine that a comprehensive, code-base–wide
review is essential - e.g., the issue spans multiple modules or exposes a systemic architectural flaw — do not proceed
with partial analysis. Instead, respond ONLY with the JSON below (and nothing else). Clearly state the reason why
you strongly feel this is necessary and ask Claude to inform the user why you're switching to a different tool:
you strongly feel this is necessary and ask the agent to inform the user why you're switching to a different tool:
{"status": "full_codereview_required",
"important": "Please use zen's codereview tool instead",
"reason": "<brief, specific rationale for escalation>"}

10
systemprompts/chat_prompt.py
View File

@@ -3,24 +3,24 @@ Chat tool system prompt
"""
CHAT_PROMPT = """
You are a senior engineering thought-partner collaborating with Claude. Your mission is to brainstorm, validate ideas,
You are a senior engineering thought-partner collaborating with another AI agent. Your mission is to brainstorm, validate ideas,
and offer well-reasoned second opinions on technical decisions when they are justified and practical.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
IF MORE INFORMATION IS NEEDED
If Claude is discussing specific code, functions, or project components that was not given as part of the context,
If the agent is discussing specific code, functions, or project components that was not given as part of the context,
and you need additional context (e.g., related files, configuration, dependencies, test files) to provide meaningful
collaboration, you MUST respond ONLY with this JSON format (and nothing else). Do NOT ask for the same file you've been
provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -34,7 +34,7 @@ SCOPE & FOCUS
and may not arise in the foreseeable future.
COLLABORATION APPROACH
1. Engage deeply with Claude's input – extend, refine, and explore alternatives ONLY WHEN they are well-justified and materially beneficial.
1. Engage deeply with the agent's input – extend, refine, and explore alternatives ONLY WHEN they are well-justified and materially beneficial.
2. Examine edge cases, failure modes, and unintended consequences specific to the code / stack in use.
3. Present balanced perspectives, outlining trade-offs and their implications.
4. Challenge assumptions constructively while respecting current design choices and goals.

6
systemprompts/codereview_prompt.py
View File

@@ -10,7 +10,7 @@ performance, maintainability, and architecture. Your task is to review the code
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -21,7 +21,7 @@ a complete and accurate review, you MUST respond ONLY with this JSON format (and
same file you've been provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -86,7 +86,7 @@ After listing issues, add:
• **Positive aspects** worth retaining
IF SCOPE TOO LARGE FOR FOCUSED REVIEW
If the codebase is too large or complex to review effectively in a single response, you MUST request Claude to
If the codebase is too large or complex to review effectively in a single response, you MUST request the agent to
provide smaller, more focused subsets for review. Respond ONLY with this JSON format (and nothing else):
{"status": "focused_review_required",
"reason": "<brief explanation of why the scope is too large>",

8
systemprompts/consensus_prompt.py
View File

@@ -4,7 +4,7 @@ Consensus tool system prompt for multi-model perspective gathering
CONSENSUS_PROMPT = """
ROLE
You are an expert technical consultant providing consensus analysis on proposals, plans, and ideas. Claude will present you
You are an expert technical consultant providing consensus analysis on proposals, plans, and ideas. The agent will present you
with a technical proposition and your task is to deliver a structured, rigorous assessment that helps validate feasibility
and implementation approaches.
@@ -14,7 +14,7 @@ analysis to make informed decisions that affect their success.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -32,7 +32,7 @@ provide thorough analysis of TECHNICAL IMPLEMENTATION details, you MUST ONLY res
Do NOT ask for the same file you've been provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -110,7 +110,7 @@ QUALITY STANDARDS
- Reference concrete examples and precedents when possible
REMINDERS
- Your assessment will be synthesized with other expert opinions by Claude
- Your assessment will be synthesized with other expert opinions by the agent
- Aim to provide unique insights that complement other perspectives
- If files are provided, reference specific technical details in your analysis
- Maintain professional objectivity while being decisive in your recommendations

26
systemprompts/debug_prompt.py
View File

@@ -4,12 +4,12 @@ Debug tool system prompt
DEBUG_ISSUE_PROMPT = """
ROLE
You are an expert debugging assistant receiving systematic investigation findings from Claude.
Claude has performed methodical investigation work following systematic debugging methodology.
Your role is to provide expert analysis based on Claude's comprehensive investigation.
You are an expert debugging assistant receiving systematic investigation findings from another AI agent.
The agent has performed methodical investigation work following systematic debugging methodology.
Your role is to provide expert analysis based on the comprehensive investigation presented to you.
SYSTEMATIC INVESTIGATION CONTEXT
Claude has followed a systematic investigation approach:
The agent has followed a systematic investigation approach:
1. Methodical examination of error reports and symptoms
2. Step-by-step code analysis and evidence collection
3. Use of tracer tool for complex method interactions when needed
@@ -18,13 +18,13 @@ Claude has followed a systematic investigation approach:
You are receiving:
1. Issue description and original symptoms
2. Claude's systematic investigation findings (comprehensive analysis)
2. The agent's systematic investigation findings (comprehensive analysis)
3. Essential files identified as critical for understanding the issue
4. Error context, logs, and diagnostic information
5. Tracer tool analysis results (if complex flow analysis was needed)
TRACER TOOL INTEGRATION AWARENESS
If Claude used the tracer tool during investigation, the findings will include:
If the agent used the tracer tool during investigation, the findings will include:
- Method call flow analysis
- Class dependency mapping
- Side effect identification
@@ -33,14 +33,14 @@ This provides deep understanding of how code interactions contribute to the issu
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
WORKFLOW CONTEXT
Your task is to analyze Claude's systematic investigation and provide expert debugging analysis back to Claude, who will
then present the findings to the user in a consolidated format.
Your task is to analyze the systematic investigation given to you and provide expert debugging analysis back to the
agent, who will then present the findings to the user in a consolidated format.
STRUCTURED JSON OUTPUT FORMAT
You MUST respond with a properly formatted JSON object following this exact schema.
@@ -50,7 +50,7 @@ IF MORE INFORMATION IS NEEDED:
If you lack critical information to proceed, you MUST only respond with the following:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -119,7 +119,7 @@ CRITICAL DEBUGGING PRINCIPLES:
4. Document your investigation process systematically for future reference
5. Rank hypotheses by likelihood based on evidence from the actual code and logs provided
6. Always include specific file:line references for exact locations of issues
7. CRITICAL: If Claude's investigation finds no concrete evidence of a bug correlating to reported symptoms,
7. CRITICAL: If the agent's investigation finds no concrete evidence of a bug correlating to reported symptoms,
you should consider that the reported issue may not actually exist, may be a misunderstanding, or may be
conflated with something else entirely. In such cases, recommend gathering more information from the user
through targeted questioning rather than continuing to hunt for non-existent bugs
@@ -129,7 +129,7 @@ When you identify specific code locations for hypotheses, include optional preci
- function_name: The exact function/method name where the issue occurs
- start_line/end_line: Line numbers from the LINE│ markers (for reference ONLY - never include LINE│ in generated code)
- context_start_text/context_end_text: Exact text from those lines for verification
- These fields help Claude locate exact positions for implementing fixes
- These fields help the agent locate exact positions for implementing fixes
REGRESSION PREVENTION: Before suggesting any fix, thoroughly analyze the proposed change to ensure it does not
introduce new issues or break existing functionality. Consider:
@@ -142,7 +142,7 @@ Your debugging approach should generate focused hypotheses ranked by likelihood,
the exact root cause and implementing minimal, targeted fixes while maintaining comprehensive documentation
of the investigation process.
Your analysis should build upon Claude's systematic investigation to provide:
Your analysis should build upon the agent's systematic investigation to provide:
- Expert validation of hypotheses
- Additional insights based on systematic findings
- Specific implementation guidance for fixes

2
systemprompts/docgen_prompt.py
View File

@@ -4,7 +4,7 @@ Documentation generation tool system prompt
DOCGEN_PROMPT = """
ROLE
You are Claude, and you're being guided through a systematic documentation generation workflow.
You're being guided through a systematic documentation generation workflow.
This tool helps you methodically analyze code and generate comprehensive documentation with:
- Proper function/method/class documentation
- Algorithmic complexity analysis (Big O notation when applicable)

14
systemprompts/planner_prompt.py
View File

@@ -13,19 +13,19 @@ efficient, and implementation-ready.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
IF MORE INFORMATION IS NEEDED
If Claude is discussing specific code, functions, or project components that was not given as part of the context,
If the agent is discussing specific code, functions, or project components that was not given as part of the context,
and you need additional context (e.g., related files, configuration, dependencies, test files) to provide meaningful
collaboration, you MUST respond ONLY with this JSON format (and nothing else). Do NOT ask for the same file you've been
provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -67,7 +67,7 @@ IF MORE INFORMATION IS NEEDED:
If you lack critical information to proceed with planning, you MUST only respond with:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["<file name here>", "<or some folder/>"]
}
@@ -92,7 +92,7 @@ FOR NORMAL PLANNING RESPONSES:
"continuation_id": "<thread_id for conversation continuity>",
"planning_complete": <true/false - set to true only on final step>,
"plan_summary": "<complete plan summary - only include when planning_complete is true>",
"next_steps": "<guidance for Claude on next actions>",
"next_steps": "<guidance for the agent on next actions>",
"previous_plan_context": "<context from previous completed plans - only on step 1 with continuation_id>"
}
@@ -101,10 +101,10 @@ PLANNING CONTENT GUIDELINES:
- Include specific actions, prerequisites, outcomes, and considerations
- When branching, clearly explain the alternative approach and when to use it
- When completing planning, provide comprehensive plan_summary
- next_steps: Always guide Claude on what to do next (continue planning, implement, or branch)
- next_steps: Always guide the agent on what to do next (continue planning, implement, or branch)
PLAN PRESENTATION GUIDELINES:
When planning is complete (planning_complete: true), Claude should present the final plan with:
When planning is complete (planning_complete: true), the agent should present the final plan with:
- Clear headings and numbered phases/sections
- Visual elements like ASCII charts for workflows, dependencies, or sequences
- Bullet points and sub-steps for detailed breakdowns

4
systemprompts/precommit_prompt.py
View File

@@ -20,7 +20,7 @@ Your feedback helps ensure this code won't cause silent regressions, developer c
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -31,7 +31,7 @@ review—and without which your analysis would be incomplete or inaccurate—you
(and nothing else). Do NOT request files you've already been provided unless their content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}

22
systemprompts/refactor_prompt.py
View File

@@ -5,15 +5,15 @@ Refactor tool system prompt
REFACTOR_PROMPT = """
ROLE
You are a principal software engineer specializing in intelligent code refactoring. You identify concrete improvement
opportunities and provide precise, actionable suggestions with exact line-number references that Claude can
opportunities and provide precise, actionable suggestions with exact line-number references that the agent can
implement directly.
CRITICAL: You MUST respond ONLY in valid JSON format. NO explanations, introductions, or text outside JSON structure.
Claude cannot parse your response if you include any non-JSON content.
The agent cannot parse your response if you include any non-JSON content.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate exact positions.
included in any code you generate. Always reference specific line numbers in your replies in order to locate
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -23,7 +23,7 @@ recommendations, you MUST respond ONLY with this JSON format (and ABSOLUTELY not
Do NOT ask for the same file you've been provided unless its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -224,7 +224,7 @@ If scope is too large and refactoring would require large parts of the code to b
CRITICAL OUTPUT FORMAT REQUIREMENTS
You MUST respond with ONLY the JSON format below. NO introduction, reasoning, explanation, or additional text.
DO NOT include any text before or after the JSON. Claude cannot parse your response if you deviate from this format.
DO NOT include any text before or after the JSON. The agent cannot parse your response if you deviate from this format.
Return ONLY this exact JSON structure:
@@ -255,12 +255,12 @@ Return ONLY this exact JSON structure:
}
],
"priority_sequence": ["refactor-001", "refactor-002"],
"next_actions_for_claude": [
"next_actions": [
{
"action_type": "EXTRACT_METHOD|SPLIT_CLASS|MODERNIZE_SYNTAX|REORGANIZE_CODE|DECOMPOSE_FILE",
"target_file": "/absolute/path/to/file.ext",
"source_lines": "45-67",
"description": "Specific step-by-step action for Claude"
"description": "Specific step-by-step action for Agent"
}
],
"more_refactor_required": false,
@@ -309,15 +309,15 @@ refactoring opportunities (typically 5-10 key changes) in the standard response
to true with an explanation.
Focus on CRITICAL and HIGH severity issues first. Include full details with refactor_opportunities, priority_sequence,
and next_actions_for_claude for the immediate changes, then indicate that additional refactoring is needed.
and next_actions for the immediate changes, then indicate that additional refactoring is needed.
Claude will use the continuation_id to continue the refactoring analysis in subsequent requests when more_refactor_required is true.
The agent will use the continuation_id to continue the refactoring analysis in subsequent requests when more_refactor_required is true.
FINAL REMINDER: CRITICAL OUTPUT FORMAT ENFORCEMENT
Your response MUST start with "{" and end with "}". NO other text is allowed.
If you include ANY text outside the JSON structure, Claude will be unable to parse your response and the tool will fail.
If you include ANY text outside the JSON structure, the agent will be unable to parse your response and the tool will fail.
DO NOT provide explanations, introductions, conclusions, or reasoning outside the JSON.
ALL information must be contained within the JSON structure itself.
Provide precise, implementable refactoring guidance that Claude can execute with confidence.
Provide precise, implementable refactoring guidance that the agent can execute with confidence.
"""

20
systemprompts/secaudit_prompt.py
View File

@@ -4,12 +4,12 @@ SECAUDIT tool system prompt
SECAUDIT_PROMPT = """
ROLE
You are an expert security auditor receiving systematic investigation findings from Claude.
Claude has performed methodical security analysis following comprehensive security audit methodology.
Your role is to provide expert security analysis based on Claude's systematic investigation.
You are an expert security auditor receiving systematic investigation findings from the agent.
The agent has performed methodical security analysis following comprehensive security audit methodology.
Your role is to provide expert security analysis based on the agent's systematic investigation.
SYSTEMATIC SECURITY INVESTIGATION CONTEXT
Claude has followed a systematic security audit approach:
The agent has followed a systematic security audit approach:
1. Security scope and attack surface analysis
2. Authentication and authorization assessment
3. Input validation and data handling security review
@@ -19,21 +19,21 @@ Claude has followed a systematic security audit approach:
You are receiving:
1. Security audit scope and application context
2. Claude's systematic security investigation findings
2. The agent's systematic security investigation findings
3. Essential files identified as critical for security assessment
4. Security issues discovered with severity classifications
5. Compliance requirements and threat level assessment
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
WORKFLOW CONTEXT
Your task is to analyze Claude's systematic security investigation and provide expert security analysis back to Claude,
who will then present the findings to the user in a consolidated format.
Your task is to analyze the agent's systematic security investigation and provide expert security analysis back to the
agent, who will then present the findings to the user in a consolidated format.
STRUCTURED JSON OUTPUT FORMAT
You MUST respond with a properly formatted JSON object following this exact schema.
@@ -43,7 +43,7 @@ IF MORE INFORMATION IS NEEDED:
If you lack critical information to proceed, you MUST only respond with the following:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -389,7 +389,7 @@ When you identify specific vulnerability locations, include optional precision f
- function_name: The exact function/method name where the vulnerability exists
- start_line/end_line: Line numbers from the LINE│ markers (for reference ONLY - never include LINE│ in generated code)
- context_start_text/context_end_text: Exact text from those lines for verification
- These fields help Claude locate exact positions for implementing security fixes
- These fields help the agent locate exact positions for implementing security fixes
REMEDIATION SAFETY AND VALIDATION:
Before suggesting any security fix, thoroughly analyze the proposed change to ensure it does not:

6
systemprompts/testgen_prompt.py
View File

@@ -10,7 +10,7 @@ in equal measure. Your mission: design and write tests that surface real-world d
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -21,7 +21,7 @@ accurate test generation, you MUST respond ONLY with this JSON format (and nothi
same file you've been provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -40,7 +40,7 @@ out-of-scope cases.
fixture layout, naming, any mocking strategy, language and tooling etc).
TEST-GENERATION STRATEGY
- If a specific test, function, class, or scenario is **explicitly** requested by Claude, focus ONLY on that specific
- If a specific test, function, class, or scenario is **explicitly** requested by the agent, focus ONLY on that specific
request and do not generate broader test coverage unless explicitly asked to do so.
- Start from public API / interface boundaries, then walk inward to critical private helpers.
- Analyze function signatures, parameters, return types, and side effects

8
systemprompts/thinkdeep_prompt.py
View File

@@ -4,12 +4,12 @@ ThinkDeep tool system prompt
THINKDEEP_PROMPT = """
ROLE
You are a senior engineering collaborator working alongside Claude on complex software problems. Claude will send you
You are a senior engineering collaborator working alongside the agent on complex software problems. The agent will send you
content—analysis, prompts, questions, ideas, or theories—to deepen, validate, or extend with rigor and clarity.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
@@ -20,7 +20,7 @@ thorough analysis, you MUST ONLY respond with this exact JSON (and nothing else)
been provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -44,7 +44,7 @@ KEY FOCUS AREAS (apply when relevant)
- Integration & Deployment: ONLY IF APPLICABLE TO THE QUESTION - external systems, compatibility, configuration, operational concerns
EVALUATION
Your response will be reviewed by Claude before any decision is made. Your goal is to practically extend Claude's thinking,
Your response will be reviewed by the agent before any decision is made. Your goal is to practically extend the agent's thinking,
surface blind spots, and refine options—not to deliver final answers in isolation.
REMINDERS

14
systemprompts/tracer_prompt.py
View File

@@ -13,19 +13,19 @@ comprehensive understanding of how software components interact and execute.
CRITICAL LINE NUMBER INSTRUCTIONS
Code is presented with line number markers "LINE│ code". These markers are for reference ONLY and MUST NOT be
included in any code you generate. Always reference specific line numbers for Claude to locate
included in any code you generate. Always reference specific line numbers in your replies in order to locate
exact positions if needed to point to exact locations. Include a very short code excerpt alongside for clarity.
Include context_start_text and context_end_text as backup references. Never include "LINE│" markers in generated code
snippets.
IF MORE INFORMATION IS NEEDED
If Claude is discussing specific code, functions, or project components that was not given as part of the context,
If the agent is discussing specific code, functions, or project components that was not given as part of the context,
and you need additional context (e.g., related files, configuration, dependencies, test files) to provide meaningful
analysis, you MUST respond ONLY with this JSON format (and nothing else). Do NOT ask for the same file you've been
provided unless for some reason its content is missing or incomplete:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["[file name here]", "[or some folder/]"]
}
@@ -70,7 +70,7 @@ IF MORE INFORMATION IS NEEDED:
If you lack critical information to proceed with tracing, you MUST only respond with:
{
"status": "files_required_to_continue",
"mandatory_instructions": "<your critical instructions for Claude>",
"mandatory_instructions": "<your critical instructions for the agent>",
"files_needed": ["<file name here>", "<or some folder/>"]
}
@@ -99,7 +99,7 @@ FOR NORMAL TRACING RESPONSES:
"continuation_id": "<thread_id for conversation continuity>",
"tracing_complete": <true/false - set to true only on final step>,
"trace_summary": "<complete trace summary - only include when tracing_complete is true>",
"next_steps": "<guidance for Claude on next investigation actions>",
"next_steps": "<guidance for the agent on next investigation actions>",
"output": {
"instructions": "<formatting instructions for final output>",
"format": "<precision_trace_analysis or dependencies_trace_analysis>",
@@ -114,10 +114,10 @@ TRACING CONTENT GUIDELINES:
- Reference exact line numbers and code snippets for evidence
- Document execution paths, call chains, or dependency relationships
- When completing tracing, provide comprehensive trace_summary
- next_steps: Always guide Claude on what to investigate next
- next_steps: Always guide the agent on what to investigate next
TRACE PRESENTATION GUIDELINES:
When tracing is complete (tracing_complete: true), Claude should present the final trace with:
When tracing is complete (tracing_complete: true), the agent should present the final trace with:
FOR PRECISION MODE:
- Vertical indented call flow diagrams with exact file:line references

2
tests/test_refactor.py
View File

@@ -45,7 +45,7 @@ class TestRefactorTool:
}
],
"priority_sequence": ["refactor-001"],
"next_actions_for_claude": [],
"next_actions": [],
},
ensure_ascii=False,
)

6
tools/analyze.py
View File

@@ -2,7 +2,7 @@
AnalyzeWorkflow tool - Step-by-step code analysis with systematic investigation
This tool provides a structured workflow for comprehensive code and file analysis.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides the CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough code examination, pattern identification, and architectural assessment before proceeding.
The tool supports complex analysis scenarios including architectural review, performance analysis,
security assessment, and maintainability evaluation.
@@ -348,7 +348,7 @@ class AnalyzeTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_analysis_summary(consolidated_findings)
context_parts.append(
f"\\n=== CLAUDE'S ANALYSIS INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
f"\\n=== AGENT'S ANALYSIS INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
)
# Add analysis configuration context if available
@@ -477,7 +477,7 @@ class AnalyzeTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Analyze-specific skip reason."""
return "Claude completed comprehensive analysis"
return "Completed comprehensive analysis locally"
def get_skip_expert_analysis_status(self) -> str:
"""Analyze-specific expert analysis skip status."""

2
tools/chat.py
View File

@@ -204,7 +204,7 @@ class ChatTool(SimpleTool):
Format the chat response to match the original Chat tool exactly.
"""
return (
f"{response}\n\n---\n\n**Claude's Turn:** Evaluate this perspective alongside your analysis to "
f"{response}\n\n---\n\nAGENT'S TURN: Evaluate this perspective alongside your analysis to "
"form a comprehensive solution and continue with the user's request and task at hand."
)

10
tools/codereview.py
View File

@@ -2,7 +2,7 @@
CodeReview Workflow tool - Systematic code review with step-by-step analysis
This tool provides a structured workflow for comprehensive code review and analysis.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides the CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough code examination, issue identification, and quality assessment before proceeding.
The tool supports complex review scenarios including security analysis, performance evaluation,
and architectural assessment.
@@ -358,7 +358,7 @@ class CodeReviewTool(WorkflowTool):
"""
Decide when to call external model based on investigation completeness.
Don't call expert analysis if Claude has certain confidence - trust their judgment.
Don't call expert analysis if the CLI agent has certain confidence - trust their judgment.
"""
# Check if user requested to skip assistant model
if request and not self.get_request_use_assistant_model(request):
@@ -380,7 +380,7 @@ class CodeReviewTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_code_review_summary(consolidated_findings)
context_parts.append(
f"\\n=== CLAUDE'S CODE REVIEW INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
f"\\n=== AGENT'S CODE REVIEW INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
)
# Add review configuration context if available
@@ -479,7 +479,7 @@ class CodeReviewTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
Code review workflow skips expert analysis when Claude has "certain" confidence.
Code review workflow skips expert analysis when the CLI agent has "certain" confidence.
"""
return request.confidence == "certain" and not request.next_step_required
@@ -516,7 +516,7 @@ class CodeReviewTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Code review-specific skip reason."""
return "Claude completed comprehensive code review with full confidence"
return "Completed comprehensive code review with full confidence locally"
def get_skip_expert_analysis_status(self) -> str:
"""Code review-specific expert analysis skip status."""

20
tools/consensus.py
View File

@@ -2,12 +2,12 @@
Consensus tool - Step-by-step multi-model consensus with expert analysis
This tool provides a structured workflow for gathering consensus from multiple models.
It guides Claude through systematic steps where Claude first provides its own analysis,
It guides the CLI agent through systematic steps where the CLI agent first provides its own analysis,
then consults each requested model one by one, and finally synthesizes all perspectives.
Key features:
- Step-by-step consensus workflow with progress tracking
- Claude's initial neutral analysis followed by model-specific consultations
- The CLI agent's initial neutral analysis followed by model-specific consultations
- Context-aware file embedding
- Support for stance-based analysis (for/against/neutral)
- Final synthesis combining all perspectives
@@ -153,7 +153,7 @@ class ConsensusTool(WorkflowTool):
"""
Consensus workflow tool for step-by-step multi-model consensus gathering.
This tool implements a structured consensus workflow where Claude first provides
This tool implements a structured consensus workflow where the CLI agent first provides
its own neutral analysis, then consults each specified model individually,
and finally synthesizes all perspectives into a unified recommendation.
"""
@@ -189,7 +189,7 @@ class ConsensusTool(WorkflowTool):
)
def get_system_prompt(self) -> str:
# For Claude's initial analysis, use a neutral version of the consensus prompt
# For the CLI agent's initial analysis, use a neutral version of the consensus prompt
return CONSENSUS_PROMPT.replace(
"{stance_prompt}",
"""BALANCED PERSPECTIVE
@@ -325,7 +325,7 @@ of the evidence, even when it strongly points in one direction.""",
Note: confidence parameter is kept for compatibility with base class but not used.
"""
if step_number == 1:
# Claude's initial analysis
# CLI Agent's initial analysis
return [
"You've provided your initial analysis. The tool will now consult other models.",
"Wait for the next step to receive the first model's response.",
@@ -416,7 +416,7 @@ of the evidence, even when it strongly points in one direction.""",
current_idx = request.current_model_index or 0
if request.step_number == 1:
# After Claude's initial analysis, prepare to consult first model
# After CLI Agent's initial analysis, prepare to consult first model
response_data["status"] = "consulting_models"
response_data["next_model"] = self.models_to_consult[0] if self.models_to_consult else None
response_data["next_steps"] = (
@@ -475,9 +475,9 @@ of the evidence, even when it strongly points in one direction.""",
"next_step_required": request.step_number < request.total_steps,
}
# Add Claude's analysis to step 1
# Add CLAI Agent's analysis to step 1
if request.step_number == 1:
response_data["claude_analysis"] = {
response_data["agent_analysis"] = {
"initial_analysis": request.step,
"findings": request.findings,
}
@@ -682,7 +682,7 @@ of the evidence, even when it strongly points in one direction.""",
"""
Customize metadata for consensus workflow to accurately reflect multi-model nature.
The default workflow metadata shows the model running Claude's analysis steps,
The default workflow metadata shows the model running Agent's analysis steps,
but consensus is a multi-model tool that consults different models. We need
to provide accurate metadata that reflects this.
"""
@@ -728,7 +728,7 @@ of the evidence, even when it strongly points in one direction.""",
}
)
# Remove the misleading single model metadata that shows Claude's execution model
# Remove the misleading single model metadata that shows Agent's execution model
# instead of the models being consulted
metadata.pop("model_used", None)
metadata.pop("provider_used", None)

8
tools/debug.py
View File

@@ -304,7 +304,7 @@ class DebugIssueTool(WorkflowTool):
"""
Decide when to call external model based on investigation completeness.
Don't call expert analysis if Claude has certain confidence - trust their judgment.
Don't call expert analysis if the CLI agent has certain confidence - trust their judgment.
"""
# Check if user requested to skip assistant model
if request and not self.get_request_use_assistant_model(request):
@@ -326,7 +326,7 @@ class DebugIssueTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_investigation_summary(consolidated_findings)
context_parts.append(
f"\n=== CLAUDE'S INVESTIGATION FINDINGS ===\n{investigation_summary}\n=== END FINDINGS ==="
f"\n=== AGENT'S INVESTIGATION FINDINGS ===\n{investigation_summary}\n=== END FINDINGS ==="
)
# Add error context if available
@@ -468,7 +468,7 @@ class DebugIssueTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
Debug tool skips expert analysis when Claude has "certain" confidence.
Debug tool skips expert analysis when agent has "certain" confidence.
"""
return request.confidence == "certain" and not request.next_step_required
@@ -501,7 +501,7 @@ class DebugIssueTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Debug-specific skip reason."""
return "Claude identified exact root cause with minimal fix requirement"
return "Identified exact root cause with minimal fix requirement locally"
def get_request_relevant_context(self, request) -> list:
"""Get relevant_context for debug tool."""

6
tools/docgen.py
View File

@@ -198,7 +198,7 @@ class DocgenTool(WorkflowTool):
"""
Docgen tool doesn't require model resolution at the MCP boundary.
The docgen tool is a self-contained workflow tool that guides Claude through
The docgen tool is a self-contained workflow tool that guides the CLI agent through
systematic documentation generation without calling external AI models.
Returns:
@@ -503,7 +503,7 @@ class DocgenTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
Docgen tool skips expert analysis when Claude has "certain" confidence.
Docgen tool skips expert analysis when the CLI agent has "certain" confidence.
"""
return request.confidence == "certain" and not request.next_step_required
@@ -536,7 +536,7 @@ class DocgenTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Docgen-specific skip reason."""
return "Claude completed comprehensive documentation analysis"
return "Completed comprehensive documentation analysis locally"
def get_request_relevant_context(self, request) -> list:
"""Get relevant_context for docgen tool."""

4
tools/listmodels.py
View File

@@ -221,7 +221,7 @@ class ListModelsTool(BaseTool):
output_lines.append(f"**Error loading models**: {str(e)}")
else:
output_lines.append("**Status**: Not configured (set OPENROUTER_API_KEY)")
output_lines.append("**Note**: Provides access to GPT-4, Claude, Mistral, and many more")
output_lines.append("**Note**: Provides access to GPT-4, O3, Mistral, and many more")
output_lines.append("")
@@ -292,7 +292,7 @@ class ListModelsTool(BaseTool):
# Add usage tips
output_lines.append("\n**Usage Tips**:")
output_lines.append("- Use model aliases (e.g., 'flash', 'o3', 'opus') for convenience")
output_lines.append("- In auto mode, Claude will select the best model for each task")
output_lines.append("- In auto mode, the CLI Agent will select the best model for each task")
output_lines.append("- Custom models are only available when CUSTOM_API_URL is set")
output_lines.append("- OpenRouter provides access to many cloud models with one API key")

16
tools/models.py
View File

@@ -17,12 +17,12 @@ class ToolModelCategory(Enum):
class ContinuationOffer(BaseModel):
"""Offer for Claude to continue conversation when Gemini doesn't ask follow-up"""
"""Offer for CLI agent to continue conversation when Gemini doesn't ask follow-up"""
continuation_id: str = Field(
..., description="Thread continuation ID for multi-turn conversations across different tools"
)
note: str = Field(..., description="Message explaining continuation opportunity to Claude")
note: str = Field(..., description="Message explaining continuation opportunity to CLI agent")
remaining_turns: int = Field(..., description="Number of conversation turns remaining")
@@ -48,7 +48,7 @@ class ToolOutput(BaseModel):
content_type: Literal["text", "markdown", "json"] = "text"
metadata: Optional[dict[str, Any]] = Field(default_factory=dict)
continuation_offer: Optional[ContinuationOffer] = Field(
None, description="Optional offer for Claude to continue conversation"
None, description="Optional offer for Agent to continue conversation"
)
@@ -56,7 +56,7 @@ class FilesNeededRequest(BaseModel):
"""Request for missing files / code to continue"""
status: Literal["files_required_to_continue"] = "files_required_to_continue"
mandatory_instructions: str = Field(..., description="Critical instructions for Claude regarding required context")
mandatory_instructions: str = Field(..., description="Critical instructions for Agent regarding required context")
files_needed: Optional[list[str]] = Field(
default_factory=list, description="Specific files that are needed for analysis"
)
@@ -75,7 +75,7 @@ class FullCodereviewRequired(BaseModel):
class FocusedReviewRequired(BaseModel):
"""Request for Claude to provide smaller, focused subsets of code for review"""
"""Request for Agent to provide smaller, focused subsets of code for review"""
status: Literal["focused_review_required"] = "focused_review_required"
reason: str = Field(..., description="Why the current scope is too large for effective review")
@@ -122,14 +122,14 @@ class RefactorOpportunity(BaseModel):
class RefactorAction(BaseModel):
"""Next action for Claude to implement refactoring"""
"""Next action for Agent to implement refactoring"""
action_type: Literal["EXTRACT_METHOD", "SPLIT_CLASS", "MODERNIZE_SYNTAX", "REORGANIZE_CODE", "DECOMPOSE_FILE"] = (
Field(..., description="Type of action to perform")
)
target_file: str = Field(..., description="Absolute path to target file")
source_lines: str = Field(..., description="Line range (e.g., '45-67')")
description: str = Field(..., description="Step-by-step action description for Claude")
description: str = Field(..., description="Step-by-step action description for CLI Agent")
class RefactorAnalysisComplete(BaseModel):
@@ -138,7 +138,7 @@ class RefactorAnalysisComplete(BaseModel):
status: Literal["refactor_analysis_complete"] = "refactor_analysis_complete"
refactor_opportunities: list[RefactorOpportunity] = Field(..., description="List of refactoring opportunities")
priority_sequence: list[str] = Field(..., description="Recommended order of refactoring IDs")
next_actions_for_claude: list[RefactorAction] = Field(..., description="Specific actions for Claude to implement")
next_actions: list[RefactorAction] = Field(..., description="Specific actions for the agent to implement")
class CodeTooLargeRequest(BaseModel):

10
tools/precommit.py
View File

@@ -2,7 +2,7 @@
Precommit Workflow tool - Step-by-step pre-commit validation with expert analysis
This tool provides a structured workflow for comprehensive pre-commit validation.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides the CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough code examination, git change analysis, and issue detection before proceeding.
The tool supports backtracking, finding updates, and expert analysis integration.
@@ -365,7 +365,7 @@ class PrecommitTool(WorkflowTool):
"""
Decide when to call external model based on investigation completeness.
Don't call expert analysis if Claude has certain confidence - trust their judgment.
Don't call expert analysis if the CLI agent has certain confidence - trust their judgment.
"""
# Check if user requested to skip assistant model
if request and not self.get_request_use_assistant_model(request):
@@ -387,7 +387,7 @@ class PrecommitTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_precommit_summary(consolidated_findings)
context_parts.append(
f"\\n=== CLAUDE'S PRE-COMMIT INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
f"\\n=== AGENT'S PRE-COMMIT INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
)
# Add git configuration context if available
@@ -485,7 +485,7 @@ class PrecommitTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
Precommit workflow skips expert analysis when Claude has "certain" confidence.
Precommit workflow skips expert analysis when the CLI agent has "certain" confidence.
"""
return request.confidence == "certain" and not request.next_step_required
@@ -522,7 +522,7 @@ class PrecommitTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Precommit-specific skip reason."""
return "Claude completed comprehensive pre-commit validation with full confidence"
return "Completed comprehensive pre-commit validation with full confidence locally"
def get_skip_expert_analysis_status(self) -> str:
"""Precommit-specific expert analysis skip status."""

12
tools/refactor.py
View File

@@ -2,7 +2,7 @@
Refactor tool - Step-by-step refactoring analysis with expert validation
This tool provides a structured workflow for comprehensive code refactoring analysis.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough code examination, refactoring opportunity identification, and quality
assessment before proceeding. The tool supports complex refactoring scenarios including
code smell detection, decomposition planning, modernization opportunities, and organization improvements.
@@ -92,7 +92,7 @@ REFACTOR_FIELD_DESCRIPTIONS = {
"Indicate your current confidence in the refactoring analysis completeness. Use: 'exploring' (starting "
"analysis), 'incomplete' (just started or significant work remaining), 'partial' (some refactoring "
"opportunities identified but more analysis needed), 'complete' (comprehensive refactoring analysis "
"finished with all major opportunities identified and Claude can handle 100% confidently without help). "
"finished with all major opportunities identified and the CLI agent can handle 100% confidently without help). "
"Use 'complete' ONLY when you have fully analyzed all code, identified all significant refactoring "
"opportunities, and can provide comprehensive recommendations without expert assistance. When files are "
"too large to read fully or analysis is uncertain, use 'partial'. Using 'complete' prevents expert "
@@ -357,7 +357,7 @@ class RefactorTool(WorkflowTool):
"""
Decide when to call external model based on investigation completeness.
Don't call expert analysis if Claude has certain confidence and complete refactoring - trust their judgment.
Don't call expert analysis if the CLI agent has certain confidence and complete refactoring - trust their judgment.
"""
# Check if user requested to skip assistant model
if request and not self.get_request_use_assistant_model(request):
@@ -383,7 +383,7 @@ class RefactorTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_refactoring_summary(consolidated_findings)
context_parts.append(
f"\\n=== CLAUDE'S REFACTORING INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
f"\\n=== AGENT'S REFACTORING INVESTIGATION ===\\n{investigation_summary}\\n=== END INVESTIGATION ==="
)
# Add refactor configuration context if available
@@ -484,7 +484,7 @@ class RefactorTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
Refactor workflow skips expert analysis when Claude has "complete" confidence.
Refactor workflow skips expert analysis when the CLI agent has "complete" confidence.
"""
return request.confidence == "complete" and not request.next_step_required
@@ -524,7 +524,7 @@ class RefactorTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Refactor-specific skip reason."""
return "Claude completed comprehensive refactoring analysis with full confidence"
return "Completed comprehensive refactoring analysis with full confidence locally"
def get_skip_expert_analysis_status(self) -> str:
"""Refactor-specific expert analysis skip status."""

8
tools/secaudit.py
View File

@@ -2,7 +2,7 @@
SECAUDIT Workflow tool - Comprehensive security audit with systematic investigation
This tool provides a structured workflow for comprehensive security assessment and analysis.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides the CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough security examination, vulnerability identification, and compliance assessment
before proceeding. The tool supports complex security scenarios including OWASP Top 10 coverage,
compliance framework mapping, and technology-specific security patterns.
@@ -351,7 +351,7 @@ class SecauditTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_security_audit_summary(consolidated_findings)
context_parts.append(
f"\n=== CLAUDE'S SECURITY INVESTIGATION ===\n{investigation_summary}\n=== END INVESTIGATION ==="
f"\n=== AGENT'S SECURITY INVESTIGATION ===\n{investigation_summary}\n=== END INVESTIGATION ==="
)
# Add security configuration context if available
@@ -566,7 +566,7 @@ class SecauditTool(WorkflowTool):
return step_data
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""Security audit workflow skips expert analysis when Claude has "certain" confidence."""
"""Security audit workflow skips expert analysis when the CLI agent has "certain" confidence."""
return request.confidence == "certain" and not request.next_step_required
def store_initial_issue(self, step_description: str):
@@ -762,7 +762,7 @@ class SecauditTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Security audit-specific skip reason."""
return "Claude completed comprehensive security audit with full confidence"
return "Completed comprehensive security audit with full confidence locally"
def get_skip_expert_analysis_status(self) -> str:
"""Security audit-specific expert analysis skip status."""

8
tools/testgen.py
View File

@@ -2,7 +2,7 @@
TestGen Workflow tool - Step-by-step test generation with expert validation
This tool provides a structured workflow for comprehensive test generation.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides the CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough code examination, test planning, and pattern identification before proceeding.
The tool supports backtracking, finding updates, and expert analysis integration for
comprehensive test suite generation.
@@ -315,7 +315,7 @@ class TestGenTool(WorkflowTool):
# Add investigation summary
investigation_summary = self._build_test_generation_summary(consolidated_findings)
context_parts.append(
f"\n=== CLAUDE'S TEST PLANNING INVESTIGATION ===\n{investigation_summary}\n=== END INVESTIGATION ==="
f"\n=== AGENT'S TEST PLANNING INVESTIGATION ===\n{investigation_summary}\n=== END INVESTIGATION ==="
)
# Add relevant code elements if available
@@ -388,7 +388,7 @@ class TestGenTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
Test generation workflow skips expert analysis when Claude has "certain" confidence.
Test generation workflow skips expert analysis when the CLI agent has "certain" confidence.
"""
return request.confidence == "certain" and not request.next_step_required
@@ -425,7 +425,7 @@ class TestGenTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Test generation-specific skip reason."""
return "Claude completed comprehensive test planning with full confidence"
return "Completed comprehensive test planning with full confidence locally"
def get_skip_expert_analysis_status(self) -> str:
"""Test generation-specific expert analysis skip status."""

4
tools/thinkdeep.py
View File

@@ -277,7 +277,7 @@ class ThinkDeepTool(WorkflowTool):
def should_skip_expert_analysis(self, request, consolidated_findings) -> bool:
"""
ThinkDeep tool skips expert analysis when Claude has "certain" confidence.
ThinkDeep tool skips expert analysis when the CLI agent has "certain" confidence.
"""
return request.confidence == "certain" and not request.next_step_required
@@ -299,7 +299,7 @@ class ThinkDeepTool(WorkflowTool):
def get_skip_reason(self) -> str:
"""Reason for skipping expert analysis."""
return "Claude expressed certain confidence in the deep thinking analysis - no additional validation needed"
return "Expressed 'certain' confidence in the deep thinking analysis - no additional validation needed"
def get_completion_message(self) -> str:
"""Message for completion without expert analysis."""

4
tools/tracer.py
View File

@@ -2,7 +2,7 @@
Tracer Workflow tool - Step-by-step code tracing and dependency analysis
This tool provides a structured workflow for comprehensive code tracing and analysis.
It guides Claude through systematic investigation steps with forced pauses between each step
It guides the CLI agent through systematic investigation steps with forced pauses between each step
to ensure thorough code examination, dependency mapping, and execution flow analysis before proceeding.
The tracer guides users through sequential code analysis with full context awareness and
@@ -545,7 +545,7 @@ class TracerTool(WorkflowTool):
def _get_rendering_instructions(self, trace_mode: str) -> str:
"""
Get mode-specific rendering instructions for Claude.
Get mode-specific rendering instructions for the CLI agent.
Args:
trace_mode: Either "precision" or "dependencies"

2
tools/version.py
View File

@@ -224,7 +224,7 @@ class VersionTool(BaseTool):
output_lines.append("git pull")
output_lines.append("```")
output_lines.append("")
output_lines.append("*Note: Restart your Claude session after updating to use the new version.*")
output_lines.append("*Note: Restart your session after updating to use the new version.*")
elif comparison == 0:
# Up to date
output_lines.append("")