Harness Kit

skill

research

Web research, multi-AI delegation, and multi-perspective validation. /research [query], /research delegate [task], /research thinktank [topic]. Use when: "search for", "look up", "research", "delegate", "get perspectives", "web search", "find out", "investigate", "introspect", "session analysis", "check readwise", "saved articles", "reading list", "highlights", "what are people saying", "X search", "social sentiment", "trending". Triggers: "search for", "look up", "research", "delegate", "get perspectives", "web search", "find out", "investigate", "introspect", "session analysis", "check readwise", "saved articles", "reading list", "highlights", "what are people saying", "X search", "social sentiment", "trending".

/research

Source: skills/research/SKILL.md

What it does

Web research, multi-AI delegation, and multi-perspective validation. /research [query], /research delegate [task], /research thinktank [topic]. Use when: "search for", "look up", "research", "delegate", "get perspectives", "web search", "find out", "investigate", "introspect", "session analysis", "check readwise", "saved articles", "reading list", "highlights", "what are people saying", "X search", "social sentiment", "trending". Triggers: "search for", "look up", "research", "delegate", "get perspectives", "web search", "find out", "investigate", "introspect", "session analysis", "check readwise", "saved articles", "reading list", "highlights", "what are people saying", "X search", "social sentiment", "trending".

Workflow role

Explore primitive in the Harness Kit operating loop.

Source contract preview

This generated excerpt gives readers the beginning of the live primitive contract before they jump to GitHub.

Evidence-backed research. The lead owns framing, source weighting, synthesis,
and residual uncertainty.

## Route

| Need | Load |
|---|---|
| broad research, comparison, architecture prior art, or discourse scan | `references/default-fanout.md` |
| `web-search`, `web-deep`, `web-news`, `web-docs` | `references/web-search.md` |
| `delegate` | `references/delegate.md` |
| `thinktank` | `references/thinktank.md` |
| `introspect` | `references/introspect.md` |
| `readwise` | `references/readwise.md` |
| `xai` | `references/xai-search.md` |
| `exemplars` | `references/exemplars.md` |

If the user names a sub-capability, load that reference. Otherwise use the
default fanout for substantive research; narrow to one source only for explicit
single-source requests or simple fact/version lookups.

## Contract

- Read the live repo first for repo facts.
- Use current external sources for drift-prone facts.
- Keep provider CLIs and web tools thin: launch, bound, record.
- Treat web search, Thinktank, provider lanes, and local grep as evidence
  inputs, not substitutes for lead synthesis.
- Separate source evidence from conclusions; cite URLs, files, commands,
  receipts, or artifacts for claims.
- Label skipped, failed, stale, in-flight, and partial sources.
- Report residual uncertainty instead of smoothing over missing evidence.

## Delegation Floor

Delegation floor applies: probe the roster first; dispatch two or more
providers for substantive work; direct solo only for mechanical, emergency,
user-forbidden, or fewer-than-two-providers cases. See
`harnesses/shared/AGENTS.md` (Roster).

Local lane guidance: Use lanes with distinct sources, methods, or model families; web search, Thinktank, and provider CLIs are evidence inputs, not substitutes for lead synthesis. Native in-thread subagents do not count toward the roster floor.
...

What to verify