fix(agents): Add PR number extraction from natural language and URLs

[rjmurillo]

Agent Context Inference Gap: PR Numbers Not Extracted from Initial Prompts

Problem

AI agents prompt users for information already provided in the initial request, violating autonomy principles.

Example:

User: "Review PR 806 comments... https://github.com/rjmurillo/ai-agents/pull/806"
Agent: "What PR number would you like me to review?"

Impact

• User Friction: Redundant clarification questions
• Autonomy Violation: Agents should infer discoverable context
• Wasted Tokens: Unnecessary back-and-forth
• Poor UX: "The whole point of this project is to have agents be independent and self-sufficient"

Root Cause

Three-Layer Problem:

1. Agent Definitions: No context inference guidance in pr-comment-responder.md (1,551 lines, zero extraction logic)
2. Missing Utilities: No shared GitHub context extraction patterns (URLs, text patterns)
3. Global Guidance: AGENTS.md lacks "extract before prompt" principle

Acceptance Criteria

• Agent extracts PR numbers from text ("PR 806", "fix(ci): Add explicit --repo flag to gh CLI commands to prevent PR context confusion #806")
• Agent parses GitHub URLs (/pull/123, /issues/456)
• Agent errors (not prompts) during autonomous execution when context is missing
• Shared utility available for all agents
• Updated AGENTS.md with context inference requirements
• Test coverage for extraction patterns

Recommended Solution

Phase 1: Quick Fix (P0, 2 hours)

Add "Phase -1: Context Inference" to pr-comment-responder skill:

# Extract from text patterns
if ($userPrompt -match 'PR #?(\d+)|pull request #?(\d+)|#(\d+)') {
    $prNumber = $matches[1] -or $matches[2] -or $matches[3]
}

# Extract from GitHub URL
if ($userPrompt -match 'github\.com/([^/]+)/([^/]+)/pull/(\d+)') {
    $owner = $matches[1]
    $repo = $matches[2]
    $prNumber = $matches[3]
}

# Error if missing during autonomous execution
if (-not $prNumber -and $autonomous) {
    throw "Cannot extract PR number from prompt. Provide explicit number."
}

Phase 2: Shared Utility (P1, 4 hours)

Create .claude/skills/github/scripts/utils/Extract-GitHubContext.ps1:

• Reusable across all agents
• Handles PR/issue text and URL patterns
• Repository extraction from URLs
• Includes Pester test coverage

Phase 3: Global Guidance (P1, 1 hour)

Update AGENTS.md with "Context Inference Requirements":

• Pattern matching table for GitHub URLs
• "Inference Before Clarification" rule
• Utility integration examples

Related

• Autonomous Execution Guardrails memory: Stricter protocols during unattended execution
• Agent Workflow Scope Discipline memory: Infer discoverable context

Evidence

• Session log: 2026-01-07-session-131.md (PR chore: remove deprecated Validate-Session.ps1 and consolidate into Validate-SessionProtocol.ps1 #810 handling - wrong PR mentioned in log, should be fix(ci): Add explicit --repo flag to gh CLI commands to prevent PR context confusion #806)
• User quote: "I don't like how it prompted for the PR number. It was in the prompt I initially gave--you should be able to intuit."

Total Effort

• P0: 2 hours
• P1: 5 hours
• Total: ~7 hours

[traycerai]

Plan

Observations

The PR comment responder agent currently lacks context inference capabilities, requiring users to re-specify PR numbers already present in their initial prompts. The agent workflow begins at Phase 0 (Memory Initialization) without any preceding extraction logic. The codebase follows PowerShell-only conventions (ADR-005), uses Pester for testing, and organizes GitHub utilities in file:.claude/skills/github/modules/GitHubCore.psm1 with scripts in file:.claude/skills/github/scripts/. No existing extraction utilities were found for parsing PR numbers from text or URLs.

Approach

Implement a three-phase solution aligned with the issue's recommended approach: (1) Add immediate context extraction to the PR comment responder agent as "Phase -1", (2) Create a reusable PowerShell utility module for GitHub context extraction, and (3) Update global agent guidance in file:AGENTS.md. This approach prioritizes quick wins while building toward reusable infrastructure, following the project's PowerShell-first architecture and existing skill patterns.

Implementation Steps

Phase 1: Quick Fix - Add Context Inference to PR Comment Responder (P0, ~2 hours)

1.1 Update PR Comment Responder Agent Definition

Files to modify:

• file:.claude/agents/pr-comment-responder.md
• file:src/claude/pr-comment-responder.md
• file:src/vs-code-agents/pr-comment-responder.agent.md
• file:src/copilot-cli/pr-comment-responder.agent.md
• file:templates/agents/pr-comment-responder.shared.md

Changes:

Add new "Phase -1: Context Inference" section immediately before Phase 0 in the workflow protocol:

### Phase -1: Context Inference (BLOCKING)

**Purpose**: Extract PR number and repository context from user prompt before any operations.

#### Step -1.1: Extract PR Number from User Input

Attempt extraction in this priority order:

1. **Direct PR number pattern**: Match `PR #123`, `PR 123`, `#123`, `pull request #123`
2. **GitHub URL pattern**: Parse `https://github.com/owner/repo/pull/123`
3. **Relative URL pattern**: Parse `/pull/123` or `pull/123`

**PowerShell extraction logic:**

```powershell

Step -1.2: Handle Missing Context

During autonomous execution (no user interaction available):

• If PR number cannot be extracted, throw error with clear message
• Do NOT prompt user - fail fast with actionable error

During interactive execution:

• If extraction fails, prompt user once for PR number
• Store extracted number for subsequent operations

Error handling:

Step -1.3: Validate Extracted Context

Before proceeding to Phase 0:

• Verify PR number is positive integer
• If repository info extracted from URL, validate format
• Log extraction method used for debugging

Proceed to Phase 0 once PR number is confirmed.

#### 1.2 Update Argument Hints

Modify the agent metadata section to include examples of valid PR references:

```markdown
**Argument Hint**: Provide PR number or URL (e.g., "Review PR #806", "https://github.com/owner/repo/pull/806", or "Review comments on pull/806")

Phase 2: Shared Utility - Create Reusable Extraction Module (P1, ~4 hours)

2.1 Create Extract-GitHubContext Utility

New file: file:.claude/skills/github/scripts/utils/Extract-GitHubContext.ps1

Function signature:

Extraction patterns to implement:

Pattern Type	Regex	Example Match
PR text	PR\s*#?(\d+)	"PR #806", "PR 806"
Issue text	Issue\s*#?(\d+)	"Issue #123"
Hash reference	#(\d+)	"#806"
Full PR URL	github\.com/([^/]+)/([^/]+)/pull/(\d+)	https://github.com/owner/repo/pull/806
Full Issue URL	github\.com/([^/]+)/([^/]+)/issues/(\d+)	https://github.com/owner/repo/issues/123
Relative PR URL	/pull/(\d+)	"/pull/806"
Relative Issue URL	/issues/(\d+)	"/issues/123"

Implementation requirements:

• Use GitHubCore module's validation functions (Test-GitHubNameValid)
• Return structured output with all extracted components
• Support pipeline input for batch processing
• Include comprehensive error messages
• Follow PowerShell best practices from file:AGENTS.md

2.2 Add Function to GitHubCore Module

File to modify: file:.claude/skills/github/modules/GitHubCore.psm1

Add Extract-GitHubContext to the module exports and include it in the function list. Place the function in a new "CONTEXT_EXTRACTION" section after the existing sections.

2.3 Create Pester Tests

New file: file:test/claude/skills/github/Extract-GitHubContext.Tests.ps1

Test coverage requirements:

Test file structure:

• Follow existing test patterns from file:tests/Invoke-CopilotAssignment.Tests.ps1
• Use BeforeAll block to load the script
• Include both pattern-based and functional tests
• Achieve 100% code coverage

2.4 Update Skill Documentation

File to modify: file:.claude/skills/github/SKILL.md (if exists) or create it

Add documentation for the new utility:

### Context Extraction

**Script**: `scripts/utils/Extract-GitHubContext.ps1`

**Purpose**: Extract PR/issue numbers and repository context from natural language or URLs.

**Usage:**
```powershell

Returns:

• Type: 'PR' or 'Issue'
• Number: Extracted number
• Owner: Repository owner (if available)
• Repo: Repository name (if available)
• Source: Extraction method used

### Phase 3: Global Guidance - Update Agent Documentation (P1, ~1 hour)

#### 3.1 Update AGENTS.md with Context Inference Requirements

**File to modify:** `file:AGENTS.md`

Add new section after "Boundaries & Constraints" (around line 256):

```markdown
## Context Inference Requirements

**Principle**: Agents MUST extract discoverable context before prompting users.

### Inference Before Clarification Rule

1. **Attempt extraction first**: Use pattern matching and parsing before asking
2. **Fail fast in autonomous mode**: Throw errors, don't prompt
3. **Prompt once in interactive mode**: If extraction fails, ask user once
4. **Log extraction method**: For debugging and improvement

### GitHub Context Extraction Patterns

Use `file:.claude/skills/github/scripts/utils/Extract-GitHubContext.ps1` for standardized extraction.

| Context Type | Pattern Examples | Extraction Method |
|--------------|------------------|-------------------|
| PR Number | "PR #123", "PR 123", "#123" | Text pattern matching |
| Issue Number | "Issue #123", "#123" | Text pattern matching |
| PR URL | `github.com/owner/repo/pull/123` | URL parsing |
| Issue URL | `github.com/owner/repo/issues/123` | URL parsing |
| Relative URL | "/pull/123", "pull/123" | Relative path parsing |

### Integration Example

```powershell

Autonomous Execution Guardrails

During autonomous execution (no user interaction):

• MUST throw descriptive errors when context missing
• MUST NOT prompt for clarification
• MUST include extraction patterns in error messages
• SHOULD log attempted extraction methods

Anti-Patterns

❌ Bad: Prompt for information already in user input

User: "Review PR 806 comments"
Agent: "What PR number would you like me to review?"

✅ Good: Extract and confirm

User: "Review PR 806 comments"
Agent: "Extracted PR #806. Proceeding with review..."

#### 3.2 Update Agent Catalog Entry

**File to modify:** `file:AGENTS.md` (Agent Catalog section, around line 512)

Update the pr-comment-responder entry:

```markdown
| **pr-comment-responder** | PR review coordinator who **extracts PR context from prompts**, gathers comment context, acknowledges every piece of feedback, and ensures systematic resolution | PR review responses, comment triage, feedback tracking |

3.3 Add Memory Reference

Create or update memory file for cross-session persistence:

File: file:.serena/memories/context-inference-patterns.md

# Context Inference Patterns

## GitHub Context Extraction

### Patterns
- PR numbers: `PR #\d+`, `PR \d+`, `#\d+`
- URLs: `github.com/owner/repo/pull/\d+`
- Relative: `/pull/\d+`

### Utility
- Script: `.claude/skills/github/scripts/utils/Extract-GitHubContext.ps1`
- Module: `GitHubCore::Extract-GitHubContext`

### Usage Stats
- Extraction success rate: Track per pattern
- Common failure modes: Document for improvement

### Related Memories
- `usage-mandatory`: Skill usage requirements
- `autonomous-execution-guardrails`: Error handling in autonomous mode

Verification Steps

Phase 1 Verification

1. Test PR comment responder with various input formats:
   • "Review PR fix(ci): Add explicit --repo flag to gh CLI commands to prevent PR context confusion #806"
   • "https://github.com/owner/repo/pull/806"
   • "fix(ci): Add explicit --repo flag to gh CLI commands to prevent PR context confusion #806"
   • "Check comments on pull request 806"
2. Verify autonomous mode throws errors (not prompts) when context missing
3. Confirm extraction logging appears in output

Phase 2 Verification

1. Run Pester tests: Invoke-Pester test/claude/skills/github/Extract-GitHubContext.Tests.ps1
2. Verify 100% test coverage
3. Test utility standalone with various inputs
4. Confirm integration with GitHubCore module

Phase 3 Verification

1. Review AGENTS.md changes for clarity and completeness
2. Verify memory file is accessible via Serena MCP
3. Confirm documentation examples are accurate
4. Test that other agents can reference the new patterns

Sequence Diagram

sequenceDiagram
    participant User
    participant Agent as PR Comment Responder
    participant Extractor as Extract-GitHubContext
    participant GitHub as GitHub API
    
    User->>Agent: "Review PR #806 comments"
    
    Note over Agent: Phase -1: Context Inference
    Agent->>Extractor: Extract-GitHubContext -InputText "Review PR #806"
    Extractor-->>Agent: {Type: PR, Number: 806, Source: TextPattern}
    
    Note over Agent: Phase 0: Memory Init
    Agent->>Agent: Load pr-comment-responder-skills memory
    
    Note over Agent: Phase 1: Context Gathering
    Agent->>GitHub: Get-PRContext -PullRequest 806
    GitHub-->>Agent: PR metadata
    
    Agent->>GitHub: Get-PRReviewers -PullRequest 806
    GitHub-->>Agent: Reviewer list
    
    Agent->>GitHub: Get-PRReviewComments -PullRequest 806
    GitHub-->>Agent: All comments
    
    Note over Agent: Phase 2-9: Process comments
    Agent->>User: "Extracted PR #806. Processing 15 comments..."

Loading

Dependencies

• PowerShell 7.4+: Required for cross-platform support
• Pester 5.6+: For test execution
• GitHub CLI 2.60+: For API operations
• GitHubCore module: Existing validation functions

Rollout Strategy

1. Phase 1 (P0): Deploy to .claude/agents/pr-comment-responder.md first for immediate impact
2. Phase 2 (P1): Create utility and tests, then update agent to use utility
3. Phase 3 (P1): Update documentation after utility is proven in production
4. Gradual adoption: Other agents can adopt the utility as needed

Success Metrics

• Zero redundant prompts: Agent never asks for PR number when present in input
• 100% test coverage: All extraction patterns covered by Pester tests
• Extraction success rate: >95% for common input formats
• Error clarity: Autonomous mode errors include extraction pattern examples

Import In IDE

🤖 Prompt for AI Agents

## Observations

The PR comment responder agent currently lacks context inference capabilities, requiring users to re-specify PR numbers already present in their initial prompts. The agent workflow begins at Phase 0 (Memory Initialization) without any preceding extraction logic. The codebase follows PowerShell-only conventions (ADR-005), uses Pester for testing, and organizes GitHub utilities in `file:.claude/skills/github/modules/GitHubCore.psm1` with scripts in `file:.claude/skills/github/scripts/`. No existing extraction utilities were found for parsing PR numbers from text or URLs.

## Approach

Implement a three-phase solution aligned with the issue's recommended approach: (1) Add immediate context extraction to the PR comment responder agent as "Phase -1", (2) Create a reusable PowerShell utility module for GitHub context extraction, and (3) Update global agent guidance in `file:AGENTS.md`. This approach prioritizes quick wins while building toward reusable infrastructure, following the project's PowerShell-first architecture and existing skill patterns.

## Implementation Steps

### Phase 1: Quick Fix - Add Context Inference to PR Comment Responder (P0, ~2 hours)

#### 1.1 Update PR Comment Responder Agent Definition

**Files to modify:**
- `file:.claude/agents/pr-comment-responder.md`
- `file:src/claude/pr-comment-responder.md`
- `file:src/vs-code-agents/pr-comment-responder.agent.md`
- `file:src/copilot-cli/pr-comment-responder.agent.md`
- `file:templates/agents/pr-comment-responder.shared.md`

**Changes:**

Add new "Phase -1: Context Inference" section immediately before Phase 0 in the workflow protocol:

```markdown
### Phase -1: Context Inference (BLOCKING)

**Purpose**: Extract PR number and repository context from user prompt before any operations.

#### Step -1.1: Extract PR Number from User Input

Attempt extraction in this priority order:

1. **Direct PR number pattern**: Match `PR #123`, `PR 123`, `#123`, `pull request #123`
2. **GitHub URL pattern**: Parse `https://github.com/owner/repo/pull/123`
3. **Relative URL pattern**: Parse `/pull/123` or `pull/123`

**PowerShell extraction logic:**

```powershell
```

#### Step -1.2: Handle Missing Context

**During autonomous execution** (no user interaction available):
- If PR number cannot be extracted, throw error with clear message
- Do NOT prompt user - fail fast with actionable error

**During interactive execution**:
- If extraction fails, prompt user once for PR number
- Store extracted number for subsequent operations

**Error handling:**

```powershell
```

#### Step -1.3: Validate Extracted Context

Before proceeding to Phase 0:
- Verify PR number is positive integer
- If repository info extracted from URL, validate format
- Log extraction method used for debugging

**Proceed to Phase 0** once PR number is confirmed.
```

#### 1.2 Update Argument Hints

Modify the agent metadata section to include examples of valid PR references:

```markdown
**Argument Hint**: Provide PR number or URL (e.g., "Review PR #806", "https://github.com/owner/repo/pull/806", or "Review comments on pull/806")
```

### Phase 2: Shared Utility - Create Reusable Extraction Module (P1, ~4 hours)

#### 2.1 Create Extract-GitHubContext Utility

**New file:** `file:.claude/skills/github/scripts/utils/Extract-GitHubContext.ps1`

**Function signature:**

```powershell
```

**Extraction patterns to implement:**

| Pattern Type | Regex | Example Match |
|--------------|-------|---------------|
| PR text | `PR\s*#?(\d+)` | "PR #806", "PR 806" |
| Issue text | `Issue\s*#?(\d+)` | "Issue #123" |
| Hash reference | `#(\d+)` | "#806" |
| Full PR URL | `github\.com/([^/]+)/([^/]+)/pull/(\d+)` | `https://github.com/owner/repo/pull/806` |
| Full Issue URL | `github\.com/([^/]+)/([^/]+)/issues/(\d+)` | `https://github.com/owner/repo/issues/123` |
| Relative PR URL | `/pull/(\d+)` | "/pull/806" |
| Relative Issue URL | `/issues/(\d+)` | "/issues/123" |

**Implementation requirements:**
- Use `GitHubCore` module's validation functions (`Test-GitHubNameValid`)
- Return structured output with all extracted components
- Support pipeline input for batch processing
- Include comprehensive error messages
- Follow PowerShell best practices from `file:AGENTS.md`

#### 2.2 Add Function to GitHubCore Module

**File to modify:** `file:.claude/skills/github/modules/GitHubCore.psm1`

Add `Extract-GitHubContext` to the module exports and include it in the function list. Place the function in a new "CONTEXT_EXTRACTION" section after the existing sections.

#### 2.3 Create Pester Tests

**New file:** `file:test/claude/skills/github/Extract-GitHubContext.Tests.ps1`

**Test coverage requirements:**

```powershell
```

**Test file structure:**
- Follow existing test patterns from `file:tests/Invoke-CopilotAssignment.Tests.ps1`
- Use `BeforeAll` block to load the script
- Include both pattern-based and functional tests
- Achieve 100% code coverage

#### 2.4 Update Skill Documentation

**File to modify:** `file:.claude/skills/github/SKILL.md` (if exists) or create it

Add documentation for the new utility:

```markdown
### Context Extraction

**Script**: `scripts/utils/Extract-GitHubContext.ps1`

**Purpose**: Extract PR/issue numbers and repository context from natural language or URLs.

**Usage:**
```powershell
```

**Returns:**
- `Type`: 'PR' or 'Issue'
- `Number`: Extracted number
- `Owner`: Repository owner (if available)
- `Repo`: Repository name (if available)
- `Source`: Extraction method used
```

### Phase 3: Global Guidance - Update Agent Documentation (P1, ~1 hour)

#### 3.1 Update AGENTS.md with Context Inference Requirements

**File to modify:** `file:AGENTS.md`

Add new section after "Boundaries & Constraints" (around line 256):

```markdown
## Context Inference Requirements

**Principle**: Agents MUST extract discoverable context before prompting users.

### Inference Before Clarification Rule

1. **Attempt extraction first**: Use pattern matching and parsing before asking
2. **Fail fast in autonomous mode**: Throw errors, don't prompt
3. **Prompt once in interactive mode**: If extraction fails, ask user once
4. **Log extraction method**: For debugging and improvement

### GitHub Context Extraction Patterns

Use `file:.claude/skills/github/scripts/utils/Extract-GitHubContext.ps1` for standardized extraction.

| Context Type | Pattern Examples | Extraction Method |
|--------------|------------------|-------------------|
| PR Number | "PR #123", "PR 123", "#123" | Text pattern matching |
| Issue Number | "Issue #123", "#123" | Text pattern matching |
| PR URL | `github.com/owner/repo/pull/123` | URL parsing |
| Issue URL | `github.com/owner/repo/issues/123` | URL parsing |
| Relative URL | "/pull/123", "pull/123" | Relative path parsing |

### Integration Example

```powershell
```

### Autonomous Execution Guardrails

During autonomous execution (no user interaction):
- **MUST** throw descriptive errors when context missing
- **MUST NOT** prompt for clarification
- **MUST** include extraction patterns in error messages
- **SHOULD** log attempted extraction methods

### Anti-Patterns

❌ **Bad**: Prompt for information already in user input
```
User: "Review PR 806 comments"
Agent: "What PR number would you like me to review?"
```

✅ **Good**: Extract and confirm
```
User: "Review PR 806 comments"
Agent: "Extracted PR #806. Proceeding with review..."
```
```

#### 3.2 Update Agent Catalog Entry

**File to modify:** `file:AGENTS.md` (Agent Catalog section, around line 512)

Update the pr-comment-responder entry:

```markdown
| **pr-comment-responder** | PR review coordinator who **extracts PR context from prompts**, gathers comment context, acknowledges every piece of feedback, and ensures systematic resolution | PR review responses, comment triage, feedback tracking |
```

#### 3.3 Add Memory Reference

Create or update memory file for cross-session persistence:

**File:** `file:.serena/memories/context-inference-patterns.md`

```markdown
# Context Inference Patterns

## GitHub Context Extraction

### Patterns
- PR numbers: `PR #\d+`, `PR \d+`, `#\d+`
- URLs: `github.com/owner/repo/pull/\d+`
- Relative: `/pull/\d+`

### Utility
- Script: `.claude/skills/github/scripts/utils/Extract-GitHubContext.ps1`
- Module: `GitHubCore::Extract-GitHubContext`

### Usage Stats
- Extraction success rate: Track per pattern
- Common failure modes: Document for improvement

### Related Memories
- `usage-mandatory`: Skill usage requirements
- `autonomous-execution-guardrails`: Error handling in autonomous mode
```

## Verification Steps

### Phase 1 Verification
1. Test PR comment responder with various input formats:
   - "Review PR #806"
   - "https://github.com/owner/repo/pull/806"
   - "/pull/806"
   - "Check comments on pull request 806"
2. Verify autonomous mode throws errors (not prompts) when context missing
3. Confirm extraction logging appears in output

### Phase 2 Verification
1. Run Pester tests: `Invoke-Pester test/claude/skills/github/Extract-GitHubContext.Tests.ps1`
2. Verify 100% test coverage
3. Test utility standalone with various inputs
4. Confirm integration with GitHubCore module

### Phase 3 Verification
1. Review AGENTS.md changes for clarity and completeness
2. Verify memory file is accessible via Serena MCP
3. Confirm documentation examples are accurate
4. Test that other agents can reference the new patterns

## Sequence Diagram

```mermaid
sequenceDiagram
    participant User
    participant Agent as PR Comment Responder
    participant Extractor as Extract-GitHubContext
    participant GitHub as GitHub API
    
    User->>Agent: "Review PR #806 comments"
    
    Note over Agent: Phase -1: Context Inference
    Agent->>Extractor: Extract-GitHubContext -InputText "Review PR #806"
    Extractor-->>Agent: {Type: PR, Number: 806, Source: TextPattern}
    
    Note over Agent: Phase 0: Memory Init
    Agent->>Agent: Load pr-comment-responder-skills memory
    
    Note over Agent: Phase 1: Context Gathering
    Agent->>GitHub: Get-PRContext -PullRequest 806
    GitHub-->>Agent: PR metadata
    
    Agent->>GitHub: Get-PRReviewers -PullRequest 806
    GitHub-->>Agent: Reviewer list
    
    Agent->>GitHub: Get-PRReviewComments -PullRequest 806
    GitHub-->>Agent: All comments
    
    Note over Agent: Phase 2-9: Process comments
    Agent->>User: "Extracted PR #806. Processing 15 comments..."
```

## Dependencies

- **PowerShell 7.4+**: Required for cross-platform support
- **Pester 5.6+**: For test execution
- **GitHub CLI 2.60+**: For API operations
- **GitHubCore module**: Existing validation functions

## Rollout Strategy

1. **Phase 1 (P0)**: Deploy to `.claude/agents/pr-comment-responder.md` first for immediate impact
2. **Phase 2 (P1)**: Create utility and tests, then update agent to use utility
3. **Phase 3 (P1)**: Update documentation after utility is proven in production
4. **Gradual adoption**: Other agents can adopt the utility as needed

## Success Metrics

- **Zero redundant prompts**: Agent never asks for PR number when present in input
- **100% test coverage**: All extraction patterns covered by Pester tests
- **Extraction success rate**: >95% for common input formats
- **Error clarity**: Autonomous mode errors include extraction pattern examples

---

Execution Information

Branch: main
Commit: dba70ca

---

💡 Tips

Supported Commands (Inside Comments)

• Use @traycerai generate to iterate on the previous version of the implementation plan.

Supported Commands (Inside Description)

• Add @traycerai ignore anywhere in the ticket description to prevent this ticket from being processed.
• Add @traycerai branch:<branch-name> anywhere in the ticket description to specify the target branch for the implementation plan.

Community

• Join our Discord Community to get help, request features, and share feedback.
• Follow us on X/Twitter for updates and announcements.

[coderabbitai]

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

• Create Plan

Examples

• Example 1
• Example 2

---

🔗 Related PRs

#229 - fix: address PR #226 review comments and labeler workflow issues [merged]
#268 - feat(copilot-synthesis): AI-powered context synthesis with thin workflow pattern [merged]
#296 - fix(workflow): ensure copilot synthesis posts comment on successful AI output [merged]
#301 - docs: autonomous PR monitoring prompt and retrospective [merged]
#534 - docs(agent-system): formalize parallel execution pattern [merged]

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[github-actions]

Product Requirements Document

Important

This PRD was automatically generated based on issue triage escalation.
Complexity Score: 8/12 (Detailed)
Escalation Criteria: feature_request,multi_phase_work,architectural_impact

About this PRD

This PRD was generated by the Explainer Agent after the issue was flagged
for detailed analysis. The document is designed to be self-contained - it can
be passed to a separate AI agent instance for implementation without additional
context.

PRD Depth Levels:

• Standard (4-6): Brief analysis, clear requirements
• Detailed (7-9): Research section, blocking questions
• Comprehensive (10-12): Phased implementation, risk analysis

---

The PRD document is complete but file creation was blocked by permissions. Here is the full PRD content:

---

PRD: GitHub Context Extraction from Natural Language and URLs

Executive Summary

AI agents prompt users for information already provided in the initial request, creating unnecessary friction and violating autonomy principles. When a user provides "Review PR 806" or a GitHub URL, agents should extract and use that context without clarification prompts.

Scope	Status	Blocker
PR number extraction from text patterns	🟢 READY	None
URL parsing for owner/repo/number	🟢 READY	None
Error behavior in autonomous execution	🟢 READY	None
Shared utility for all agents	🟢 READY	None
AGENTS.md context inference guidance	🟢 READY	None
Test coverage	🟢 READY	None

Verdict: READY. All requirements can be implemented with existing patterns.

Problem Statement

Current State

Component	Current Behavior	Location
pr-comment-responder skill	No context extraction logic	.claude/skills/pr-comment-responder/SKILL.md
pr-comment-responder agent	Prompts for PR number even when provided	src/claude/pr-comment-responder.md
GitHub skill	Scripts require explicit -PullRequest parameter	.claude/skills/github/scripts/
GitHubCore.psm1	No URL/text parsing utilities	.claude/skills/github/modules/GitHubCore.psm1
AGENTS.md	No "extract before prompt" principle	AGENTS.md

Gap Analysis

1. Missing Extraction Logic: No utility to parse GitHub context from "PR 806" or full URLs
2. No Inference Requirement: Agent docs do not mandate context inference before prompting
3. Inconsistent Error Behavior: Agents prompt during autonomous execution when they should error
4. Duplicated Effort: Each agent would implement extraction independently without shared utility

User Impact

Impact Type	Description
User Friction	Redundant clarification questions waste user time
Autonomy Violation	Agents should infer discoverable context
Token Waste	Each clarification round uses ~200 tokens
Poor UX	Contradicts project goal of self-sufficient agents

Proposed Solution

Phase 1 (4 hours): Create Extract-GitHubContext.ps1 utility with URL and text pattern parsing, test coverage

Phase 2 (2 hours): Update pr-comment-responder with "Phase -1: Context Inference"

Phase 3 (1 hour): Add context inference requirements to AGENTS.md

Total Effort: ~7 hours

Functional Requirements

FR-1: URL Parsing (P0)

• Parse github.com/owner/repo/pull/123 → Owner, Repo, Number, Type
• Handle with/without protocol and enterprise domains
• Extract multiple references from input

FR-2: Text Pattern Parsing (P0)

• Parse "PR 806", "PR fix(ci): Add explicit --repo flag to gh CLI commands to prevent PR context confusion #806", "fix(ci): Add explicit --repo flag to gh CLI commands to prevent PR context confusion #806", "pull request 806", "issue 123"
• Case insensitive matching

FR-3: Error Behavior (P1)

• Interactive mode: return null on missing context
• Autonomous mode: throw error on missing context

Technical Design

Create Extract-GitHubContext.ps1 in .claude/skills/github/scripts/utils/ following existing PowerShell patterns from Get-PRContext.ps1 and GitHubCore.psm1.

Success Metrics

Metric	Target
User clarification prompts for provided context	0 occurrences
Test coverage for extraction utility	>= 80%
Time to implement	<= 7 hours

---

_{Generated by AI PRD Generation | View Workflow}