← All skills · Automation · Research Brief Compiler

Research Brief Compiler — finished output

Runs inClaude Code Codex Cursor

Muse

Dot Grok Bot

Cue OpenClaw

Hermes

Install guidesClaude Code Codex Cursor Muse

Needs:Your collected notes: research.txt (Topic and Question lines, then one Note line per item — id | source pointer | observation — one Claim line per assertion — claim text | note ids — and one Open line per unanswered question).

Automation Premium

Research Brief Compiler

Turn collected notes into a research brief with a claims table, evidence pointers, and open questions: each claim graded supported (2+ notes), single-source (1 note), or unsupported (no notes), and every claim's evidence traced back to the source pointer you recorded.

Version 1.0.0

Subscribe — All Skills, one subscription.No exclusions. Curated, verified, and continuously updated.

Get this Skill with Premium See plans

Reads one plain-text file you write — research.txt, with a Topic line and a Question line, plus Note lines (id | source pointer | observation), Claim lines (claim text | note ids), and Open lines (open questions) — and produces research-brief.md: the topic and research question first, then the claims table (claim, support level, evidence pointers), then the open questions, then the notes index. Support is a count of cited notes — supported means 2 or more, single-source means exactly 1, unsupported means none — never a judgement that a claim is true. Local-only, standard library only; a claim citing a note id that was never listed is dropped and named, and a brief with no claims or no open questions says so — the brief is a document, it publishes and cites nothing on your behalf.

Good fit Synthesizing notes you collected on a topic into one honest brief Checking which of your assertions actually have evidence behind them Handing a brief to someone else with its evidence traceable in one step

Similar Skills

Premium

Email Triage Automation

Read full threads
Sort into 3 buckets
Draft in your voice
Approve & send

Premium

Smart Scheduling Automation

Gather constraints
Check all calendars
Offer local times
Hold or invite

Free

Deadline-Driven Follow-Up

Capture the cutoff
Watch for stalls
Nudge at 3 points
Close the loop

Premium

Durable Bulk Execution

Shard the list
Checkpoint each item
Idempotent runs
Verify completion

View all skills →

What you get

A claims table with a support level on every claim — supported, single-source, or unsupported
Evidence pointers listed verbatim from the notes each claim cites
Open questions listed as-is, so the brief shows what is still unanswered
Every gap, input error, and warning named in the report

Good fit

Synthesizing notes you collected on a topic into one honest brief
Checking which of your assertions actually have evidence behind them
Handing a brief to someone else with its evidence traceable in one step

Not for

Judging whether a claim is true — support counts cited notes, nothing more
Fetching or inventing sources — evidence is only what you recorded
Publishing or citing anything — the brief is a document, not a submission

Sample output

An example of what this Skill produces. Sample content, shown for illustration.

Research Brief Compiler report — sample

Claims: 3 | Supported: 1 | Single-source: 1 | Unsupported: 1 | Open questions: 2 | Notes: 4 | Gaps: 1 | Errors: 0 | Warnings: 0

Supported: small teams merge PRs faster with AI coding agents — N1 (Blog — blog.example.com/ai-agents-shipping) and N2 (Interview — Priya, Nov 2026). Single-source: a human review step is still required before merge — N3 (Docs — docs.example.com/agents). Unsupported: AI agents cut total engineering headcount — no notes cited, listed not treated as a finding. Open: do the gains hold for large codebases, and what does “faster” cost in review quality?

Gaps: 1 claim cites no notes.

About this Skill

You say

Turn these collected notes into a research brief — what do the notes actually support, what rests on a single source, and what is still open?

How to install

Copy the install prompt into your agent: Read https://skillu.dev/install/agent.md, then install the Research Brief Compiler skill (research-brief-compiler).
Or download the Skill package from this page after launch and place its folder in your agent’s skills directory.
Call it with a slash command, e.g. /research-brief-compiler, and describe what you want in plain words.

FAQ

How is a claim's support level decided?

From the notes you cite, nothing else: supported means 2 or more distinct notes, single-source means exactly 1, unsupported means none. Two notes saying the same thing make a claim supported in this brief; they do not make it correct — the notes themselves could be wrong, outdated, or from the same original source, and that judgement is yours.

What happens to a claim that cites a note I never listed?

It is dropped and named in the input errors — the script never invents a source to back a claim. Add the missing Note line with its source pointer and re-run, and the claim will be graded from the notes you actually collected.

Get Premium access Browse more skills