Caches codebase exploration into persistent research documents before implementation begins.
npx skills add https://github.com/arndvs/ctrlshft --skill research-arndvs
Please help me install this Agent Skill. Skill: research Source: https://github.com/arndvs/ctrlshft/tree/main/skills/research Command: npx skills add https://github.com/arndvs/ctrlshft --skill research-arndvs
working/research/<topic>.md with sections for summary, architecture, constraints, dependencies, open questions, and recommendations. - Lifecycle Management: Reuses research files under 7 days old, re-validates stale ones, and promotes lasting findings to docs/research/ before cleanup. - Use Case: Before building a new authentication system, run research to map existing auth code paths, external dependencies, and API constraints, then hand the document to the implementation conversation. ## Quick Start Ask the agent to research the authentication flow before building the new login feature.High-intent search queries and answers about installing and using this skill.
Invoke the research skill with your topic, and it decomposes the investigation into 3-6 areas explored by parallel sub-agents. Findings are synthesized into a structured markdown file at working/research/<topic>.md covering architecture, constraints, and recommendations.
Run research when architecture or API choices are unclear, when exploration would consume over 20% of context, or when multiple conversations will work on the same area. It prevents re-exploration and preserves implementation capacity.
Research files under 7 days old can be used as-is if the topic matches. Older files require re-validation by spot-checking 2-3 key claims, and should be regenerated after major refactors or dependency changes.
Research files are working documents and should be deleted after the feature ships. If the synthesis has lasting value, promote it to docs/research/<topic>.md before deleting the working copy.
Yes, include the research file via @-references when starting a new conversation. If an active plan exists, reference both files together so the fresh conversation picks up remaining work with full context.
parcadeiDocument existing codebase components and interactions into structured Markdown reports.
anombyte93Expands TaskMaster tasks with parallel research agents and writes findings back to tasks.json.
dxosMaintains durable TASKS.md ledgers and a project registry for multi-session work tracking.
Fandry96Compress agent conversation history into structured summaries with artifact tracking.