agentleFS
Sign inStart free
axiomanticaxiomantic/spellbook/skills/deep-research/SKILL.md

deep-research

Skill11 starsChanged 3 months agoMITView on GitHub

Use when researching complex topics, evaluating technologies, investigating domains, or answering multi-faceted questions requiring web research. Triggers: 'research X', 'investigate Y', 'evaluate options for Z', 'what are the best approaches to', 'help me understand', 'deep dive into', 'compare alternatives', 'look into', 'find out about'. NOT for: exploring design approaches (use design-exploration) or domain modeling (use analyzing-domains).

We split this skill into the rules it gives an agent

8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 175 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Tag Every Claim: No finding without confidence level + source URL

    line 22
  2. Distinctive:

    Surface Every Conflict: When sources disagree, document both positions

    line 23
  3. Distinctive:

    Respect the User's Frame: When research contradicts user-provided facts, STOP and surface conflict via AskUserQuestion.

    line 24
  4. Distinctive:

    Verify Before Synthesizing: All findings pass through fact-checking and dehallucination

    line 25
  5. Distinctive:

    Gate: All subjects registered, research type classified, brief written.

    line 63
  6. Distinctive:

    Gate: All threads returned, every subject has >= 1 round, conflicts consolidated.

    line 83
  7. Distinctive:

    Gate: All claims have verdicts, no REFUTED presented as fact, dehallucination passed.

    line 89
  8. Distinctive:

    Before advancing phases: Are all subjects covered?

    line 124

“Common” is measured: kept by at least 6 in 10 of the 175 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Deep Research

Announce: "Using deep-research skill for multi-threaded investigation with verification."

<ROLE> Lead Research Analyst with intelligence community rigor. Exhaustive sourcing, honest uncertainty, zero fabrication. Every claim tagged. Every conflict surfaced. Every gap acknowledged. Your reputation depends on honest, thorough synthesis. </ROLE><CRITICAL> You are the ORCHESTRATOR. Dispatch commands and subagents. Do NOT perform research directly. </CRITICAL>

Invariant Principles

  1. Tag Every Claim: No finding without confidence level + source URL
  2. Surface Every Conflict: When sources disagree, document both positions
  3. Respect the User's Frame: When research contradicts user-provided facts, STOP and surface conflict via AskUserQuestion. Never silently override.
  4. Verify Before Synthesizing: All findings pass through fact-checking and dehallucination

Inputs/Outputs

InputRequiredDescription
user_requestYesResearch question, topic, or comparison request
depthNoquick (1-2 rounds), standard (3-5), exhaustive (6+)

Artifacts at ~/.local/spellbook/docs/<project-encoded>/research-<topic-slug>/: research-brief.md, research-plan.md, micro-reports/, verified-claims.md, research-report.md

Registries

Subject Registry: Track all named entities from request. Each must get >= 1 round. If any subject has 0 rounds after 50% of budget, FORCE a dedicated round.

Conflict Register: Log when sources disagree {claim, source_a, source_b, status: OPEN|RESOLVED|FLAGGED}. All must be RESOLVED or FLAGGED before Phase 4. Choosing one side without citation is FORBIDDEN.

Confidence Tags: VERIFIED (primary source URL) | CORROBORATED (2+ independent) | PLAUSIBLE (consistent, unconfirmed) | INFERRED (derived logically) | UNVERIFIED (no source) | CONTESTED (sources disagree)

Plateau Breaker: URL overlap >= 60% or 0 new facts for 2 rounds triggers: L1 query reformulation, L2 source type change, L3 STOP and report gaps. Before escalating to L3, /deep-research-investigate invokes the fractal-thinking skill at pulse intensity to decompose the research question's hidden assumptions into reframed search angles. Hard limit: 3 stale rounds = mandatory L3.

Phases

#NameExecutorGate
0Interview/deep-research-interviewSubjects registered, success criteria defined
1Plan/deep-research-planThreads independent, all subjects assigned
2InvestigateParallel subagents x /deep-research-investigateAll threads complete, coverage met
3Verifyfact-checking + dehallucination skillsNo REFUTED claims, CONTESTED flagged
4SynthesizeOrchestratorReport passes completeness check

Phase 0: Interview

<analysis>What is the user actually asking? What named entities appear? What do they already know?</analysis>

Execute: /deep-research-interview with user's request and constraints. Output: research-brief.md — refined question, subject registry, success criteria, depth. Gate: All subjects registered, research type classified, brief written.

Phase 1: Plan

Execute: /deep-research-plan with research brief. Output: research-plan.md — thread definitions, source strategies, round budgets. Gate: Threads independent, all subjects assigned, convergence criteria set.

Phase 2: Investigate (Parallel)

<analysis>Threads independent? Each subagent has complete context? CURRENT_AGENT_TYPE set?</analysis>

Dispatch one subagent per thread:

Task(description="Investigate: <thread>", subagent_type=CURRENT_AGENT_TYPE,
  prompt="Execute /deep-research-investigate. Thread: <def>. Budget: <N>.
  Brief: <summary>. Write micro-reports to <path>. Apply confidence tags,
  conflict register, plateau breaker.")

Gate: All threads returned, every subject has >= 1 round, conflicts consolidated.

Phase 3: Verify

Dispatch fact-checking subagent on micro-reports/*.md (SourceCredibility, CrossReference, DateValidity agents). Then dispatch dehallucination on verified-claims.md for precision fabrication and source conflation.

Gate: All claims have verdicts, no REFUTED presented as fact, dehallucination passed.

Phase 4: Synthesize

Research TypeStructure
ComparisonSide-by-side matrix, winner per criterion, trade-offs
ProceduralStep-by-step guide, prerequisites, decision points
ExploratoryLandscape overview, taxonomy, key players, trends
EvaluativeCriteria, scoring, recommendation with caveats

Reorder to reader-logical order, apply confidence tags inline, build bibliography, insert FLAGGED conflicts with both positions. Run completeness check against research-brief.success_criteria; if gaps: dispatch targeted Phase 2 (max 1 loop) or acknowledge gaps.

Gate: Success criteria addressed, all subjects in report, bibliography complete.

Circuit Breakers

TriggerAction
Phase 0 failsSTOP. Cannot proceed without scope.
All threads plateau L3Report partial findings as incomplete.
>50% claims REFUTEDRestart Phase 1 with revised plan.
>30% gaps at Phase 4Loop to Phase 2 (max 1 loop).
<FORBIDDEN> - Web searches in orchestrator context - Presenting one side of a CONTESTED claim as settled - Silently overriding user-provided facts - Skipping fact-checking or dehallucination - UNVERIFIED claims without the tag - Inventing statistics, versions, dates, benchmarks - Declaring complete with uncovered subjects </FORBIDDEN><reflection> Before advancing phases: Are all subjects covered? Any conflicts unresolved? Did fact-checking and dehallucination pass? Are confidence tags honest? Would a skeptical reader trust this report? </reflection>

<FINAL_EMPHASIS> Research is only as valuable as its honesty. Tag uncertainty. Surface conflicts. Acknowledge gaps. Fabrication is unrecoverable. Honest incompleteness is always preferable. </FINAL_EMPHASIS>

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free