independent-research_skill

Overview

This skill enforces research-first behavior before asking factual questions or making technical recommendations. It spins up parallel checks and web searches to verify tools, versions, configs, and known issues so any suggestion is evidence-backed. Use it to avoid guessing, stale memory, or “lazy” questions that the system can answer itself.

How this skill works

Before asking the user anything factual, the skill runs local probes (which, --version, config file reads), parallel web subagents (search docs, changelogs, GitHub issues, community answers), and synthesizes findings. If the prompt indicates a premature decision or a lazy question, the skill replaces the question with concrete commands, verified results, or links to authoritative sources. It only asks the user when a true preference or business constraint cannot be determined programmatically.

When to use it

• About to ask a factual question that could be answered programmatically (installations, versions, configs).
• Preparing to propose a fix, migration, or diagnostic that depends on current docs or known issues.
• Choosing between tools, versions, or approaches where community consensus or compatibility matters.
• Diagnosing an unfamiliar error — before recommending a change.
• Any time you suspect your memory may be stale or breaking changes could exist.

Best practices

• Run local checks first: which, tool --version, read config files, inspect package manifests.
• Spin parallel web subagents: one for official docs, one for changelogs/migration notes, one for GitHub/issue trackers, one for community Q&A.
• Search the exact error message in quotes and capture the most relevant issue or doc link before deciding.
• Never ask a factual question you can answer yourself; ask only about preferences, priorities, or inaccessible context.
• When presenting options, cite the compatibility matrix or recent changelog entries discovered during research.

Example use cases

• User reports build failure; skill runs local env checks and websearches the exact build error, then returns vetted fixes and links to matching GitHub issues.
• Before recommending a package upgrade, skill fetches the package changelog and migration guide and reports breaking changes found.
• When asked to pick between two frameworks, skill launches subagents to compare recent docs, benchmarks, and community discussions and summarizes evidence for each choice.
• Debugging CI: skill checks CI config, runner environment, recent commits, and searches for matching CI errors in upstream repos.