talont-org/review-approach

Review Approach Skill

Research modern solutions, approaches, and strategies relevant to the issues or features proposed in a report or plan. Uses web search subagents to gather external perspective and surface options the team may not have considered.

When to Use

• User says "review approach", "review approaches", or "research solutions"
• User wants to validate a proposed direction against current industry practice
• User has a plan or report and wants to explore what modern solutions exist
• After an investigation or plan, before committing to an approach

Critical Constraints

NEVER:

• Modify any source code files
• Create files outside {{AUTOSKILLIT_TEMP}}/review-approach/ directory
• Run subagents in the background (run_in_background: true is prohibited)

ALWAYS:

• Use subagents with web search for parallel research
• Use model: "sonnet" when spawning all subagents via the Task tool
• Keep findings concise and actionable
• Present options with trade-offs
• Make recommendations based on technical merit and project fit
• Tie research back to the specific problem context
• Include source URLs for all referenced material
• After writing the review file, emit the absolute path as a structured output token as your final output. Resolve the relative temp/review-approach/... save path to absolute by prepending the full CWD:

  review_path = /absolute/cwd/temp/review-approach/{filename}.md
  

  This token is MANDATORY — the pipeline cannot proceed without it.

Workflow

Step 1: Extract Research Targets

From the report, plan, or conversation context, identify the core problems and proposed features that need external research. Break them into distinct research topics.

Step 2: Launch Parallel Web Search Subagents

Spawn general-purpose subagents (with web search) for each research topic. Each subagent should investigate:

• What modern solutions exist for this problem class
• How mature projects and frameworks approach it
• Recent developments, libraries, or patterns worth considering
• Known pitfalls and trade-offs of common approaches

Tailor the search queries to the specific technologies and constraints of the project.

Step 3: Synthesize

Consolidate subagent findings into a concise review. For each research topic:

• What exists: The relevant modern approaches found
• Trade-offs: Strengths and weaknesses in the context of this project
• Relevance: How each option relates to the proposed direction

Drop anything that doesn't meaningfully inform the decision.

Step 4: Write Review

Save to: {{AUTOSKILLIT_TEMP}}/review-approach/review_approach_{topic}_{YYYY-MM-DD_HHMMSS}.md (relative to the current working directory)

# Approach Review: {Topic}

**Date:** {YYYY-MM-DD}
**Source:** {Name of the report/plan being reviewed}

## Context
{Brief statement of the problem and what was proposed}

## Research Findings

### {Research Topic 1}
{What modern solutions exist, trade-offs, relevance to this project}

**Sources:**
- [{title}]({url})

### {Research Topic 2}
{...}

## Recommendations
{What approaches to pursue and why, based on the research}

## Key Takeaways
{Concise bullets — what matters most for the decision at hand}

After saving the review file, emit the structured output token as the very last line of your text output:

IMPORTANT: Emit the structured output tokens as literal plain text with no markdown formatting on the token names. Do not wrap token names in **bold**, *italic*, or any other markdown. The adjudicator performs a regex match on the exact token name — decorators cause match failure.

review_path = {absolute_path_to_review_file}