Project Documentation Enforcement Protocol

A comprehensive protocol for GitHub Copilot that enforces strict adherence to project-specific documentation through the doc-bot MCP server, preventing generic responses and ensuring all code generation and project inquiries are validated against actual project rules.

What This Skill Does

This skill creates a multi-phase enforcement system that:

Detects the presence of doc-bot MCP server and activates strict documentation-first mode

Classifies all user requests into specific categories requiring different tool interactions

Mandates tool calls before generating responses for code generation, project inquiries, and documentation access

Blocks responses that rely on general AI knowledge when project documentation exists

Validates all responses against project rules and documentation

Provides automatic error recovery and response regeneration when violations occur

Instructions

PHASE 1: ENVIRONMENT DETECTION

**Before processing any request:**

1. Check if the project contains a `doc-bot/` directory

2. Verify if the doc-bot MCP server is available (either `doc-bot` or `@afterxleep/doc-bot`)

3. If both conditions are met:

- Activate strict documentation enforcement mode

- Disable general knowledge fallbacks for project-specific queries

- Set project documentation as the sole authoritative source

PHASE 2: REQUEST CLASSIFICATION

**Classify every user request into one of these categories:**

#### CLASS A: CODE_GENERATION

**Trigger keywords**: write, create, implement, build, add, code, function, class, component, method, develop, generate, make, refactor, modify, update, fix, debug

**Required action**: MUST call `check_project_rules` tool with task description

**Blocking**: YES - cannot proceed without tool response

**Timeout**: 30 seconds

#### CLASS B: PROJECT_INQUIRY

**Trigger keywords**: how, what, why, architecture, approach, pattern, feature, authentication, database, api, testing

**Context keywords**: this project, this codebase, here, our, current project

**Required action**: MUST call `search_documentation` tool with extracted keywords

**Blocking**: YES - cannot proceed without tool response

**Timeout**: 30 seconds

#### CLASS C: DOCUMENTATION_DISCOVERY

**Trigger keywords**: documentation, available, capabilities, help, rules, standards, what exists, docs, guide

**Required action**: MUST call `get_global_rules` tool

**Blocking**: YES - cannot proceed without tool response

**Timeout**: 30 seconds

#### CLASS D: DOCUMENT_ACCESS

**Trigger keywords**: read, show, get content, full document, complete, entire

**Context**: Previous search results are present

**Required action**: SHOULD call `read_specific_document` tool with filename

**Blocking**: NO - can proceed if file not found

**Timeout**: 15 seconds

**Special Note**: For file-specific queries (e.g., "working on src/utils.js"), use `search_documentation` with both the file name and context keywords rather than a separate tool.

PHASE 3: EXECUTION ENFORCEMENT

**Pre-Response Validation:**

1. Classify the user's request using the classification matrix

2. If classification is NOT "UNRELATED":

- Execute the mandatory tool for that classification

- If tool returns an error, provide error recovery response

- If tool succeeds, inject tool result into response context

- Mark tool result as authoritative source

- Disable general knowledge fallback

3. Do NOT generate response until tool execution completes

**Response Constraint Rules:**

If you attempt to use general knowledge when authoritative source exists: STOP and use only authoritative source

If code generation is detected and project rules haven't been checked: HALT and emit error "Project rules validation required before code generation"

All project-specific answers MUST come from tool results, not training data

PHASE 4: QUALITY ASSURANCE

**Response Validation (after generating):**

1. Check if response contradicts tool result

- If YES: Regenerate response using ONLY tool result

2. Check if response lacks tool attribution

- If YES: Append attribution showing source document/rule

3. Validate response compliance with project rules

- If FAIL: Regenerate with corrections

**Failure Detection & Recovery:**

If a response was generated WITHOUT executing required tools:

- Log violation: "Tool bypass detected"

- Immediately regenerate response with proper tool execution

- Prepend warning: "Corrected response using project documentation:"

PHASE 5: BEHAVIORAL CONSTRAINTS

**ABSOLUTE PROHIBITIONS (Never do these):**

Generate code without validating against project rules

Answer project-specific questions from general AI knowledge

Override tool responses with training data

Skip tool calls for "simple" questions when project context exists

**MANDATORY BEHAVIORS (Always do these):**

Execute tool calls BEFORE generating responses

Explicitly attribute tool results in responses

Prioritize project documentation over general knowledge

Acknowledge errors and provide corrections

Verification Examples

**Test these scenarios to verify proper behavior:**

1. **Documentation discovery:**

- User: "What documentation is available?"

- Expected: Immediate call to `get_global_rules()`, no general response

2. **Project inquiry:**

- User: "How do I implement auth?"

- Expected: Immediate call to `search_documentation("authentication")`, no generic auth patterns

3. **Code generation:**

- User: "Create a new component"

- Expected: Immediate call to `check_project_rules("create component")`, no standard React/Vue/etc. patterns

4. **File-specific work:**

- User: "Working on src/utils.js"

- Expected: Call to `search_documentation("src/utils.js utils")`, context-aware suggestions

**Expected Behavior**: Tool execution within 2 seconds of request classification, zero responses from general knowledge.

**Failure Indicators**: Any response without tool execution, generic programming advice, standard framework suggestions without project context.

Constraints

Requires doc-bot MCP server to be installed and available

Requires `doc-bot/` directory to exist in project root

Tool timeouts may require fallback handling (error recovery)

Only applies to project-specific requests; general programming questions remain unrestricted

Project Documentation Enforcement Protocol

Project Documentation Enforcement Protocol

What This Skill Does

Instructions

PHASE 1: ENVIRONMENT DETECTION

PHASE 2: REQUEST CLASSIFICATION

PHASE 3: EXECUTION ENFORCEMENT

PHASE 4: QUALITY ASSURANCE

PHASE 5: BEHAVIORAL CONSTRAINTS

Verification Examples

Constraints

Reviews (0)