All skills

Escalation workflow for stubborn bugs. Use when a bug persists after multiple fix attempts, you've tried several approaches, or you're stuck.

Use this Skill: https://skilld.dev/gh/nielsmadan/agentic-coding/hard-fix

This session only. Nothing lands on disk.

referencestemplates.md

≈731 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Hard Fix — Agent Prompts & Templates

Phase 2: Agent Task Prompts

Research Agent

research-tech {library_if_any} {error_or_symptom}

Focus on: known bugs, breaking changes, similar issues others faced

Debug Agent

debug-log {problem_area}

Add comprehensive logging to trace the exact execution path and state

History Agent

review-history {affected_files_or_area}

Look for: recent changes, when it last worked, who touched it, past similar issues

Library Source Agent

Subagent type: general-purpose
Prompt:
---
Investigate the source code of libraries involved in this issue:

Problem: {description}
Libraries involved: {library_names}

1. Find the library source code:
   - node_modules/{library}/ for JS/TS
   - Look for .dart files in pub cache for Flutter
   - site-packages/{library}/ for Python
   - vendor/ or go modules for Go

2. Locate the relevant functions/classes being used

3. Read the actual implementation and look for:
   - Undocumented behavior or edge cases
   - Default values that might cause issues
   - Error handling that swallows errors
   - Race conditions or timing assumptions
   - Version-specific behavior

4. Check if our usage matches what the library expects

Return findings about how the library actually works vs how we're using it.
---

Second Opinion Agent

second-opinion

Problem: {description}
Tried: {list of attempted fixes}
Symptoms: {what's happening}

What are we missing?

Phase 3: Synthesis Examples

BAD synthesis (superficial):

Root Cause: The API call is failing.
Evidence: Got a 500 error in the logs.
Fix: Add a try-catch.

GOOD synthesis (investigative):

Root Cause: Race condition between auth token refresh and API call.
Evidence:
- Debug logs: Token refresh starts at T+0, API call at T+50ms, refresh completes T+200ms
- History: Started after PR #234 moved token refresh to background
- Library source: axios doesn't queue requests during refresh by default
- Research: Known issue axios#4193, recommended fix is axios-auth-refresh
Fix: Add request queuing during token refresh using axios-auth-refresh interceptor.

The difference: superficial stops at symptoms, good traces to mechanism.


Phase 5: Presentation Template

## Hard Fix Analysis: {problem}

**Root Cause** (Confidence: High/Medium/Low): {mechanism, not symptom}

**Evidence**: {one key finding per source}

**Recommended Fix**: {specific steps}

**Alternatives**: {if main fix fails}

**Validation**: {how to verify}

Phase 7: Log Template

Write to docs/log/YYYY-MM-DD-{Issue}.md:

# {Issue Title}

**Date:** {date} | **Area:** {files/components}

## Problem & Symptoms
{what was happening}

## Root Cause
{what was actually wrong - the mechanism}

## Solution
{what fixed it, with code if relevant}

## Prevention
{how to avoid in future}

Source: SKILL.md on GitHub

2 warnings6mo4 checks · Risk SAFE
  • Gen Agent Trust Hub6mo

    The skill provides a structured and heavyweight debugging workflow for resolving persistent bugs. It utilizes a systematic process involving local file searches, parallel investigation using specialized agents, and a mandatory validation step before logging verified solutions. No malicious behavior, obfuscation, or unauthorized data exfiltration patterns were identified.

  • Socket6mo

    No alerts

  • Snyk6mo

    Risk: MEDIUM · 1 issue

  • Runlayer6mo

    1/1 file flagged

Signed by skilld at e8b8a29. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last week
argument-hint
<description of the persistent problem>
effort
max

README badge

README badge for nielsmadan/agentic-coding/hard-fix