Research Skill

research-skill

Persistent project-scoped store for deep research findings, with progressive disclosure and contrarian-pass investigation.

---

What this is

A Claude Code and Codex skill that gives you a persistent, project-scoped store for deep research findings.

Stop re-researching the same topics across sessions. Stop polluting conversation context with raw web search dumps. The skill maintains a structured local knowledge base under /.research/, looks it up before fetching the web, and uses progressive disclosure to load only what's actually needed.

---

Built for compaction and large-research recall

Long Claude Code sessions run out of context. /compact summarizes older turns and drops the rest, so findings from a deep research thread evaporate and the next question re-triggers the same web searches.

This skill makes the data layer outlive the chat. Research written today survives /compact, /clear, IDE restarts, and machine moves. The next session reads INDEX.md first (a tiny dispatcher), matches the topic, and pulls only the matched entry's ## Summary section into context. The full body stays on disk until you actually need it.

Loading tiers, cheapest first:

Tier	Loads	Approx tokens	When
1	INDEX.md	100 to 500	Every retrieval
2	Entry's ## Summary only	50 to 200	When INDEX shows a match
3	Full FINDINGS.md body	500 to 3000	When the summary doesn't cover it

Heavy research artifacts become cheap to recall: you only pay for the tier you need.

---

What's distinctive

• Project-scoped, not global. Each repo has its own research store, kept private (gitignored by default).
• Progressive disclosure. Index, then summary, then full body, in that order. Most lookups never load the full entry.
• Conflict-handling history. When findings change, old claims move to a ## Discarded approaches table with reasons; never silently overwritten. Prevents re-trying refuted approaches.
• Subagent-isolated investigation. Heavy web research can run in a separate subagent: an Opus-class model in Claude Code, or GPT-5.5 xhigh in Codex when subagents are explicitly authorized. Your main context stays clean. Hosts without subagents run the same phases inline; the skill states the fallback explicitly.
• Async where supported. In Claude Code, the investigation subagent runs in background mode (run_in_background: true) so the conversation stays interactive while research happens. In Codex, the plugin investigates inline unless the user explicitly authorizes subagents.
• Cognitive phases. Decompose, Gather, Validate, Contrarian pass, Insight extraction, Synthesize. The contrarian pass actively searches for "why this is wrong" rather than confirming; the insight pass forces causal and comparative claims that go beyond restating facts.
• Two depths: research and deep research. Fresh questions are triaged at retrieval: quick (3 to 6 searches, minutes) for narrow lookups, deep (the full 6-phase walk) for comparisons and anything costly to get wrong. Your wording overrides the triage in both directions; a quick entry is stored with depth: quick and upgrades through a deep merge later.
• Small-model tested. The ambiguous-question rule and the storage rule exist because Haiku-class agents actually broke without them; the retrieval stop rule was then verified holding on the same tier. The