Agent Skills: Web Research: {topic}

Comprehensive multi-wave web research with strategic source selection. Gathers information from official docs, community resources, and advanced sources. Use for deep technical research, API documentation, best practices. Triggers: research web, deep research, comprehensive research, find documentation.

UncategorizedID: doodledood/codex-workflow/research-web

Install this agent skill to your local

pnpm dlx add-skill https://github.com/doodledood/codex-workflow/tree/HEAD/skills/research-web

Skill Files

Browse the full folder contents for research-web.

Download Skill

Loading file tree…

skills/research-web/SKILL.md

Skill Metadata

Name
research-web
Description
"Comprehensive multi-wave web research with strategic source selection. Gathers information from official docs, community resources, and advanced sources. Use for deep technical research, API documentation, best practices. Triggers: research web, deep research, comprehensive research, find documentation."

User request: $ARGUMENTS

Comprehensive multi-wave web research. Strategically selects sources, evaluates credibility, cross-references findings, and synthesizes actionable recommendations.

Research log: /tmp/research-web-{YYYYMMDD-HHMMSS}-{topic-kebab-case}.md - external memory for all findings.

Phase 1: Research Planning

1.1 Parse Request

Extract from $ARGUMENTS:

  • Primary question: What are we trying to learn?
  • Context: Why do we need this? What decision does it inform?
  • Constraints: Version requirements, platform limitations, etc.

If vague or empty: Ask user to clarify the specific question.

1.2 Create Research Log

Path: /tmp/research-web-{YYYYMMDD-HHMMSS}-{topic-kebab-case}.md

# Web Research: {topic}
Started: {timestamp}

## Research Question
{Primary question}

## Context
{Why needed, what decision it informs}

## Initial Hypotheses
(populated in 1.3)

## Wave Results
(populated during research)

## Source Registry
(populated during research)

## Conflicts
(populated when sources disagree)

## Synthesis
(populated in Phase 4)

1.3 Form Initial Hypotheses

Before searching, document what you expect to find:

## Initial Hypotheses

### H1: {Expected answer}
- Confidence: Low | Medium | High
- Would falsify: {what would prove this wrong}

### H2: {Alternative possibility}
- Confidence: Low | Medium | High
- Would falsify: {what would prove this wrong}

Why hypotheses first: Prevents confirmation bias. Search for evidence both supporting AND refuting.

1.4 Create Todo List

- [ ] Wave 1: Official documentation
- [ ] Wave 2: Community resources
- [ ] Wave 3: Advanced/specialized sources
- [ ] Cross-reference and resolve conflicts
- [ ] Synthesize findings

Phase 2: Strategic Wave Research

Wave Selection Strategy

| Wave | Source Type | Purpose | When to Use | |------|-------------|---------|-------------| | 1 | Official docs, GitHub repos | Canonical behavior, API specs | Always start here | | 2 | Stack Overflow, tutorials, blogs | Practical examples, common issues | After official gaps identified | | 3 | Academic papers, expert blogs, niche forums | Edge cases, advanced techniques | When waves 1-2 insufficient |

2.1 Wave 1: Official Sources

Priority sources:

  1. Official documentation sites
  2. GitHub repositories (README, docs/, wiki)
  3. Official API references
  4. Release notes and changelogs

Search queries:

  • {topic} official documentation
  • {topic} github
  • {technology} {feature} docs

Document each source:

### Source: {title}
**URL**: {url}
**Type**: Official Documentation | GitHub | API Reference
**Date**: {publication/update date}
**Credibility**: High (official)

**Key findings**:
- {Finding 1}
- {Finding 2}

**Evidence**:
> "{Relevant quote}"

**Supports/Refutes**: H1, H2

2.2 Wave 2: Community Resources

Priority sources:

  1. Stack Overflow (recent, accepted answers)
  2. Technical blogs (reputable authors)
  3. Tutorial sites
  4. Discussion forums

Search queries:

  • {topic} example
  • {topic} tutorial
  • {topic} best practices
  • {technology} {error message} (if troubleshooting)

Credibility indicators:

  • High: Accepted answer, author is maintainer/contributor, multiple upvotes
  • Medium: Popular answer, detailed explanation, includes code
  • Low: Old answer, no engagement, no sources

2.3 Wave 3: Advanced Sources

Only proceed if waves 1-2 leave gaps.

Sources:

  • Academic papers (for algorithms, protocols)
  • Expert technical blogs
  • Conference talks
  • Specialized forums/communities
  • Source code analysis

Search queries:

  • {topic} paper
  • {topic} deep dive
  • {topic} internals

Phase 3: Research Loop

Memento Loop

For each wave:

  1. Mark wave todo in_progress
  2. Execute searches
  3. Evaluate and document each source
  4. Write findings immediately to research log
  5. Update hypothesis status
  6. Mark wave todo completed
  7. Decide: proceed to next wave or sufficient?

NEVER proceed without writing findings — research log is external memory.

3.1 Source Credibility Evaluation

For each source, assess:

| Factor | High | Medium | Low | |--------|------|--------|-----| | Authority | Official, maintainer | Expert, educator | Anonymous, unknown | | Recency | <6 months | <2 years | >2 years | | Verification | Multiple sources agree | Some corroboration | Single source | | Depth | Shows understanding | Practical focus | Surface level |

3.2 Update Hypotheses

As findings come in:

### H1: {hypothesis}
- Status: CONFIRMED | REFUTED | UNCERTAIN
- Confidence: Low | Medium | High
- Evidence: {summary}
- Sources: {list}

3.3 Handle Conflicts

When sources disagree:

## Conflicts

### {Topic of disagreement}
**Position A**: {claim}
- Sources: {list}
- Strength: {why credible}

**Position B**: {claim}
- Sources: {list}
- Strength: {why credible}

**Resolution**:
- {Your assessment based on source authority, recency, consensus}
- OR "Unresolved - present both options to user"

Conflict resolution priority:

  1. Official docs > community
  2. Recent > old
  3. Multiple sources > single source
  4. Code/specs > opinions

3.4 Wave Completion Criteria

Stop wave when:

  • Primary question answered with high confidence
  • Multiple authoritative sources agree
  • Diminishing returns (same info repeating)

Proceed to next wave when:

  • Gaps remain in understanding
  • Only low-credibility sources found
  • Conflicting information needs resolution

Phase 4: Synthesis

4.1 Refresh Context

Read the full research log before synthesizing.

4.2 Cross-Reference Findings

Verify key claims appear in multiple independent sources:

## Cross-Reference Matrix

| Finding | Official | Community | Expert |
|---------|----------|-----------|--------|
| {Finding 1} | ✓ source | ✓ source | - |
| {Finding 2} | ✓ source | ✓ source | ✓ source |
| {Finding 3} | - | ✓ source | - |

4.3 Write Synthesis

## Synthesis

### Answer to Primary Question
{Direct answer}
**Confidence**: High | Medium | Low
**Based on**: {N} sources, {consensus level}

### Key Findings
1. {Finding with source citations}
2. {Finding with source citations}
3. {Finding with source citations}

### Hypothesis Resolution
- H1: {CONFIRMED/REFUTED} — {evidence}
- H2: {CONFIRMED/REFUTED} — {evidence}

### Caveats and Limitations
- {What we couldn't verify}
- {Where information was sparse}
- {Time-sensitivity concerns}

### Recommendations
1. {Primary recommendation with rationale}
2. {Alternative approach and when to use}
3. {What to avoid and why}

### Sources by Authority

**Official**:
- {Source with URL}

**Community (High credibility)**:
- {Source with URL}

**Other**:
- {Source with URL}

4.4 Present Summary

## Research Complete

**Question**: {Primary question}
**Confidence**: High | Medium | Low
**Waves completed**: {N}
**Sources consulted**: {N}

### Answer
{Concise direct answer}

### Key Findings
- {Finding 1}
- {Finding 2}
- {Finding 3}

### Recommendations
1. {What to do}
2. {What to avoid}

### Caveats
- {Key limitation}

### Top Sources
- {Most authoritative source}
- {Second source}

**Full research log**: /tmp/research-web-{...}.md

Guidelines

DO

  • Document hypotheses BEFORE searching
  • Record ALL sources, even unhelpful ones
  • Note publication dates
  • Cross-reference important claims
  • Acknowledge uncertainty
  • Update log after EACH source

DON'T

  • Trust single sources for important claims
  • Ignore publication dates
  • Present opinions as facts
  • Skip synthesis step
  • Make claims without citations
  • Cherry-pick confirming sources

Source Authority Hierarchy

  1. Official docs — How things are supposed to work
  2. Source code — How things actually work
  3. GitHub issues — Real problems and workarounds
  4. Stack Overflow — Community solutions (CHECK DATES!)
  5. Technical blogs — Opinions and tutorials (verify claims)
  6. General articles — Background only

Time Sensitivity

| Domain | Information Half-life | |--------|----------------------| | Web frameworks | 6-12 months | | Languages | 1-2 years | | CS fundamentals | 5+ years | | Security | 3-6 months | | Cloud services | 6-12 months |

Always note when information might be outdated.

Edge Cases

| Case | Action | |------|--------| | No relevant sources | Note in log, ask user for alternative terms | | All sources outdated | Note prominently, recommend verifying with latest docs | | Conflicting authoritative sources | Present both with evidence, let user decide | | Topic too broad | Ask user to narrow scope | | Paywalled content | Note limitation, search for alternatives |