Persistent project-scoped knowledge base for deep research findings with progressive disclosure and contrarian-pass investigation.
research-skill is easy to set up with trust notes worth reviewing. Check agent compatibility and use-case fit before adding it to your workflow.
Repository setup guidance
npx skills add hec-ovi/research-skillRun the command in your terminal.
Confirm that the skill files were added to your agent workspace.
Check the README requirements before invoking the skill in your agent.
This tool stores research findings locally in your project, so you don't have to re-search the same topics across sessions. It uses a tiered loading system to keep context usage low, and includes a contrarian pass that actively looks for evidence against your conclusions.
research-skill is a skill for Claude Code and other SKILL.md-compatible code CLIs that provides a persistent, project-scoped knowledge base for deep research findings. It stores structured research data under `<project>/.research/`, preventing re-researching the same topics across sessions and reducing context pollution from raw web search dumps. The skill uses progressive disclosure: it first loads a lightweight INDEX.md (100-500 tokens), then an entry's summary (50-200 tokens), and only loads the full FINDINGS.md body (500-3000 tokens) when needed. This design survives `/compact`, `/clear`, IDE restarts, and machine moves. Key features include conflict-handling history (old claims move to a 'Discarded approaches' table with reasons), subagent-isolated investigation (heavy research runs in a separate subagent to keep main context clean), async background mode in Claude Code, and a contrarian pass that actively searches for 'why this is wrong' rather than confirming. The skill follows cognitive phases: Decompose, Gather, Validate, Contrarian pass, Synthesize.
Looks usable, but maintenance, license, or security notes deserve a closer look.
Last commit was about 82 days ago.
5 GitHub stars indicate community interest.
0 open issues signal maintenance load.
MIT license detected.
Maintain a persistent research store for a large codebase to avoid re-searching the same topics.
Conduct deep research on a complex technical problem with progressive disclosure to manage context.
Use contrarian-pass investigation to challenge assumptions and find flaws in proposed solutions.
Collaborate across sessions by sharing a project-scoped knowledge base that survives compaction.
Run heavy web research in a subagent while keeping the main conversation interactive.
Agent skills can change coding-agent behavior; review every SKILL.md and referenced script before installing.
Check shell, network, file-system, credential, and API-key requirements before running skill workflows.
The skill stores research data locally in the project directory; ensure sensitive data is not committed if gitignored.
Subagent investigation may consume additional API tokens; monitor usage.
5
Stars
0
Forks
0
Issues
MIT
License
A public repository of example skills for Claude, demonstrating how to create reusable instructions and resources for specialized tasks.
Adversarial AI bug hunter with auto-fix skill for Claude Code, Cursor, Codex CLI, GitHub Copilot CLI, Kiro CLI, Opencode, Pi Coding Agent, and more. Multi-agent pipeline finds security vulnerabilities, logic errors, and runtime bugs — then fixes them autonomously on a safe branch.
Persistent project-scoped knowledge base for deep research findings with progressive disclosure and contrarian-pass investigation.
Developers using Claude Code or Codex who need persistent research memory across sessions., Technical leads managing large projects who want to avoid re-researching the same topics., Researchers and engineers conducting deep investigations with a need for structured, tiered recall.
The setup section provides repository-level starting guidance, not a guarantee of an independently verified installation. Check the official README and release notes.
Agent skills can change coding-agent behavior; review every SKILL.md and referenced script before installing. Check shell, network, file-system, credential, and API-key requirements before running skill workflows.
Similar or complementary options to evaluate include anthropic-skills, bug-hunter, context7.
Jun 7, 2026
4 security/trust notes recorded.
Setup difficulty is 2/5.
Up-to-date code documentation for LLMs and AI code editors, eliminating outdated or hallucinated API references.