Scientific Research Team Setup#

Deploy 4 agents: literature, hypothesis, experiment, and report.

Roles#

RoleHostnameSkillsPurposeliterature<prefix>-literaturepilot-discover, pilot-archive, pilot-stream-dataSearches databases, builds literature reviewshypothesis<prefix>-hypothesispilot-task-router, pilot-priority-queue, pilot-datasetGenerates and ranks testable hypothesesexperiment<prefix>-experimentpilot-task-router, pilot-audit-log, pilot-metricsDesigns protocols, runs experiments, validates resultsreport<prefix>-reportpilot-share, pilot-announce, pilot-webhook-bridgeCompiles findings into structured reports

Setup Procedure#

Step 1: Ask the user which role this agent should play and what prefix to use.

Step 2: Install the skills for the chosen role:

bash

# For literature:
clawhub install pilot-discover pilot-archive pilot-stream-data
# For hypothesis:
clawhub install pilot-task-router pilot-priority-queue pilot-dataset
# For experiment:
clawhub install pilot-task-router pilot-audit-log pilot-metrics
# For report:
clawhub install pilot-share pilot-announce pilot-webhook-bridge

Step 3: Set the hostname and write the manifest to ~/.pilot/setups/scientific-research-team.json.

Step 4: Tell the user to initiate handshakes with the peers for their role.

Manifest Templates Per Role#

literature#

json

{
  "setup": "scientific-research-team", "role": "literature", "role_name": "Literature Agent",
  "hostname": "<prefix>-literature",
  "skills": {
    "pilot-discover": "Search scientific databases (PubMed, arXiv, Semantic Scholar).",
    "pilot-archive": "Store retrieved papers and extracted findings for reference.",
    "pilot-stream-data": "Stream structured literature summaries to hypothesis agent."
  },
  "data_flows": [
    { "direction": "send", "peer": "<prefix>-hypothesis", "port": 1002, "topic": "literature-synthesis", "description": "Structured reviews with gaps and citations" }
  ],
  "handshakes_needed": ["<prefix>-hypothesis"]
}

hypothesis#

json

{
  "setup": "scientific-research-team", "role": "hypothesis", "role_name": "Hypothesis Agent",
  "hostname": "<prefix>-hypothesis",
  "skills": {
    "pilot-task-router": "Route hypotheses to appropriate experimental methodologies.",
    "pilot-priority-queue": "Rank hypotheses by novelty, feasibility, and impact.",
    "pilot-dataset": "Access existing datasets for preliminary pattern analysis."
  },
  "data_flows": [
    { "direction": "receive", "peer": "<prefix>-literature", "port": 1002, "topic": "literature-synthesis", "description": "Literature reviews with identified gaps" },
    { "direction": "send", "peer": "<prefix>-experiment", "port": 1002, "topic": "research-hypothesis", "description": "Ranked hypotheses with proposed protocols" }
  ],
  "handshakes_needed": ["<prefix>-literature", "<prefix>-experiment"]
}

experiment#

json

{
  "setup": "scientific-research-team", "role": "experiment", "role_name": "Experiment Agent",
  "hostname": "<prefix>-experiment",
  "skills": {
    "pilot-task-router": "Manage experiment execution queues and resource allocation.",
    "pilot-audit-log": "Log all experimental parameters, runs, and raw results.",
    "pilot-metrics": "Track statistical measures -- p-values, confidence intervals, effect sizes."
  },
  "data_flows": [
    { "direction": "receive", "peer": "<prefix>-hypothesis", "port": 1002, "topic": "research-hypothesis", "description": "Hypotheses to test" },
    { "direction": "send", "peer": "<prefix>-report", "port": 1002, "topic": "experiment-result", "description": "Results with statistical validation" }
  ],
  "handshakes_needed": ["<prefix>-hypothesis", "<prefix>-report"]
}

report#

json

{
  "setup": "scientific-research-team", "role": "report", "role_name": "Report Writer",
  "hostname": "<prefix>-report",
  "skills": {
    "pilot-share": "Distribute final reports to collaborators and reviewers.",
    "pilot-announce": "Broadcast publication announcements to the research network.",
    "pilot-webhook-bridge": "Push reports to external platforms (preprint servers, journals)."
  },
  "data_flows": [
    { "direction": "receive", "peer": "<prefix>-experiment", "port": 1002, "topic": "experiment-result", "description": "Experimental results to compile" },
    { "direction": "send", "peer": "external", "port": 443, "topic": "research-report", "description": "Published reports via secure channel" }
  ],
  "handshakes_needed": ["<prefix>-experiment"]
}

Data Flows#

• literature -> hypothesis : literature synthesis with gaps and citations (port 1002)
• hypothesis -> experiment : ranked research hypotheses with protocols (port 1002)
• experiment -> report : experimental results with statistical analysis (port 1002)
• report -> external : published research reports (port 443)

Workflow Example#

bash

# On literature -- publish synthesis:
pilotctl --json publish <prefix>-hypothesis literature-synthesis '{"topic":"transformer scaling laws","papers_reviewed":47,"gaps":["sparse attention + MoE interaction unknown"]}'
# On hypothesis -- publish ranked hypothesis:
pilotctl --json publish <prefix>-experiment research-hypothesis '{"hypothesis_id":"H-031","statement":"Flash attention + MoE achieves superlinear scaling","novelty_score":0.87}'
# On experiment -- publish validated results:
pilotctl --json publish <prefix>-report experiment-result '{"hypothesis_id":"H-031","status":"partially_supported","p_value":0.003,"effect_size":"medium"}'
# On report -- announce publication:
pilotctl --json publish <prefix>-report research-report '{"title":"Flash Attention + MoE: Superlinear Scaling","status":"ready_for_review"}'

Dependencies#

Requires pilot-protocol skill, pilotctl binary, clawhub binary, and a running daemon.