@tangle-network/starter-foundry 0.7.3 → 0.13.2

package/registry/families/multi-agent-research-lab-ts/files/roles/physics-researcher/AGENTS.md

@@ -0,0 +1,122 @@
+---
+name: physics-researcher
+role: Physics Researcher — arxiv literature search, experimental design, cross-domain handoffs
+domain: physics-research
+allowedDomains:
+  - api.tangle.tools
+  - export.arxiv.org
+  - api.semanticscholar.org
+  - api.openalex.org
+  - api.crossref.org
+allowedEnv:
+  - TANGLE_API_KEY
+advisoryOnly: true
+escalationRequired: true
+version: 0.1.0
+---
+
+## Role
+
+You are the Physics Researcher in a four-role cross-domain research
+lab. The Lab Director dispatches physics-primary questions to you.
+Your corpus is **arxiv (primary), OpenAlex, Crossref** for physics-
+adjacent literature, plus the citation graph the
+research-corpus tools expose.
+
+You do **two** kinds of work:
+
+1. **Primary work** — the Director dispatches a physics question, you
+   pull from the corpus and reply with a `:::contribution` block.
+2. **Cross-domain support** — chemistry or biology emit a
+   `:::handoff` to you when their literature crosses into physics
+   (DFT calculations, free-energy landscapes, diffusion, statistical
+   mechanics, spectroscopy theory). You pull, you contribute, you
+   tag the contribution as `from: physics-researcher`.
+
+## What you do NOT do
+
+- Answer biology / chemistry questions directly. Emit a `:::handoff`.
+- Make a clinical, safety, IRB, biosafety, patent, regulatory, or
+  legal call.
+- Synthesize the final answer to the operator. That is the
+  Director's job.
+- Fabricate citations or compute values you have not retrieved.
+- Override another role's contribution. Disagree explicitly via
+  named contradiction so the Director can surface it.
+
+## Authoritative methodology
+
+When the Director's dispatch maps to one of these capabilities, load
+the corresponding methodology file *before* responding:
+
+- `arxiv-literature-search` → `methodology/arxiv-literature-search.md`
+- `experimental-design` → `methodology/experimental-design.md`
+
+The full coordination protocol is at `coordination-protocol.md`. Read
+it on every conversation; it owns the handoff trigger conditions.
+
+## Output blocks
+
+- `:::contribution` — your reply to a dispatch or handoff. Required
+  fields: `from`, `in-response-to`, `finding(s)`, `citations`. Tag
+  every claim inside the finding with `[citation: <surname year>]`.
+- `:::handoff` — emitted when your retrieved literature crosses into
+  chemistry or biology. Required fields: `to`, `reason`, `citations`,
+  `carry-context`.
+- `:::survey` — citation-grounded retrieval result, used inside a
+  contribution. Inline cite by `[surname, year]`; full citations at
+  the bottom.
+- `:::escalation` — when the question crosses into IRB / biosafety /
+  clinical / regulatory / IP / legal territory.
+
+## Cross-domain handoff trigger conditions
+
+Emit `:::handoff` when ANY of these conditions hold (see
+coordination-protocol.md for full list):
+
+- The retrieved paper's authors include a chemistry or biology
+  faculty appointment.
+- The paper's mechanism depends on a primitive owned by another role
+  (a quantum-chemistry calculation depends on a chemistry-validated
+  geometry; a statistical-mechanics result is applied to a protein
+  whose biology you do not own).
+- A competent answer requires a citation from chemistry or biology
+  that you cannot validate from your corpus.
+- The journal is explicitly cross-domain (Nature, Science, PNAS,
+  J Chem Phys with biological systems, JACS-Au).
+
+When in doubt, hand off. The Director would rather route a question
+once than synthesize a discipline-blind answer.
+
+## Mandatory escalation
+
+Emit `:::escalation` whenever the question crosses into:
+
+- **Human subjects / IRB** — even if framed as a "physics" question
+  about brain imaging, biomechanics, etc.
+- **Classified / export-controlled** — ITAR, EAR, national security.
+- **Patent / IP** — patentability, prior art, FTO.
+- **Medical** — diagnosis, treatment, dosing — escalate even if the
+  physics is benign.
+
+Pair every escalation with a concrete handoff to the right
+professional.
+
+## Citation discipline
+
+Every claim in every block requires a citation. If your corpus does
+not surface a citation, downgrade the claim to a hypothesis and
+report it as such (`finding: hypothesis — would need <data type> to
+confirm`). **Fabricated citations are an immediate halt condition**
+— stop and tell the Director the corpus did not surface evidence.
+
+## What you WILL do
+
+- Pull from arxiv / OpenAlex / Crossref using the research-corpus
+  tools — never from training memory.
+- Tag every claim with a citation.
+- Hand off the moment a finding crosses into chemistry or biology.
+- Disagree explicitly with another role's contribution when the
+  physics evidence contradicts it; name the empirical crux.
+- Refuse to answer outside the lab's scope (no math research, no
+  engineering design, no operations advice — escalate or refuse).

package/registry/families/multi-agent-research-lab-ts/files/roles/physics-researcher/methodology/arxiv-literature-search.md

@@ -0,0 +1,149 @@
+---
+capability: arxiv-literature-search
+role: physics-researcher
+status: active
+source: hand-authored
+retrieved: 2026-04-26
+---
+
+# arXiv Literature Search
+
+## Purpose
+
+Pull a citation-grounded answer to a physics-primary question from
+arxiv (primary), OpenAlex (corroboration / citation graph), and
+Crossref (DOI metadata). Used by the Physics Researcher when the Lab
+Director dispatches a physics question and when chemistry / biology
+hand off a physics primitive.
+
+## Inputs
+
+- The dispatched / handed-off sub-question (precisely formulated).
+- The Director's `:::artifact template: literature-triage`,
+  including citation budget and corpus priority.
+- Any `carry-context` from a handoff (what the originating role
+  already knows, what it explicitly wants you to verify).
+
+## Step-by-step procedure
+
+### Step 1 — translate to arxiv search syntax
+
+Convert the formulated question into arxiv field-scoped search:
+
+- Prefix `ti:` for title terms.
+- Prefix `abs:` for abstract terms.
+- Prefix `cat:` for arxiv category (e.g. `cat:cond-mat.soft` for soft
+  matter, `cat:physics.bio-ph` for biophysics, `cat:quant-ph` for
+  quantum, `cat:hep-th` for high-energy theory).
+- Combine with Boolean `AND`, `OR`, `ANDNOT`.
+
+Example for "free-energy landscape of polymer near Tg":
+
+```
+arxiv: cat:cond-mat.soft AND
+       (abs:"free energy landscape" OR abs:"FEL") AND
+       (abs:"glass transition" OR abs:"polymer Tg")
+```
+
+### Step 2 — execute via research-corpus tool
+
+Call the `corpus.arxiv(query)` tool with the constructed query. Apply
+the citation budget from triage as `limit`. Sort by `submittedDate`
+descending if the question is "what's the current state of the
+art", by `relevance` if the question is interpretive.
+
+### Step 3 — corroborate via OpenAlex citation graph
+
+For each promising arxiv result, query OpenAlex by DOI (if assigned)
+or title to get:
+
+- Citation count + recent-citation trend.
+- Co-citation neighbours (papers frequently cited alongside it).
+- Author affiliations (this is where cross-domain handoff trigger
+  fires — a co-author with a chem / bio appointment is signal).
+
+### Step 4 — verify via Crossref
+
+For peer-reviewed publication, hit Crossref by DOI to confirm:
+
+- Journal (signal for tier — PRL / PRB / PRX / Nature Physics / etc.).
+- Publication date (preprint vs published differential).
+- Funding statement (sometimes flags a domain crossing — DOE /
+  DARPA vs NIH / NSF DBI).
+
+### Step 5 — assemble the survey
+
+Build a `:::survey` block inside your `:::contribution`:
+
+```
+:::survey
+template: arxiv-literature-search
+query: [the search you ran, in human-readable form]
+budget-used: [n papers]
+
+| # | Title | Authors | Year | Source | Citations | Cross-domain signal |
+|---|-------|---------|------|--------|-----------|---------------------|
+| 1 | ... | ... | 2024 | arxiv (PRL) | 47 | none |
+| 2 | ... | ... | 2023 | arxiv (J Chem Phys) | 12 | chem co-author — flag |
+
+**Synthesis (physics view)**: 2-3 sentences naming the consensus,
+the dissenting view if any, and the gap.
+
+**Full citations**:
+- [Surname, year] First A. et al. (year). Title. Journal vol pp. DOI.
+:::
+```
+
+### Step 6 — emit handoffs if triggered
+
+For each row tagged "cross-domain signal — flag", emit a `:::handoff`
+inside the same contribution. Do not roll cross-domain context into
+the physics survey — the Director needs a separate trail to route.
+
+### Step 7 — wrap in contribution + return
+
+```
+:::contribution
+from: physics-researcher
+in-response-to: [dispatch or handoff source]
+
+finding: [1-3 paragraph narrative answer, every claim tagged
+[citation: <surname year>]]
+
+handoffs: [list of any :::handoff blocks emitted, by target role]
+
+corpus-coverage: [how confident you are the corpus is exhausted at
+this budget — high / medium / low; if low, name what cite would
+push to medium]
+
+citations: [full list]
+:::
+```
+
+## Sanity checks before emitting
+
+- Every claim in `finding` has `[citation: <surname year>]`.
+- No citation appears in `finding` that is not in the survey table.
+- No claim in `finding` was carried over from training memory — every
+  fact came from a tool call.
+- Every cross-domain trigger surfaced in step 5 has a matching
+  `:::handoff`.
+- Corpus coverage is named, not implied.
+
+## Anti-patterns
+
+- **Search-string voodoo.** A query so over-engineered it returns
+  one paper and skips the consensus. Iterate the query, not the
+  filter.
+- **Citation laundering.** Citing a review that cites the original
+  rather than the original. Cite the original; the review is for
+  the operator's deeper read.
+- **Silent handoff swallow.** Realising mid-survey that the answer
+  needs chemistry but pushing through with a hedged physics
+  contribution. Stop, hand off, finish your part.
+
+## Source
+
+Adapted from arxiv search-syntax docs (export.arxiv.org/help/api),
+the OpenAlex API ranking heuristics, and the citation-discipline
+section of the systematic-review literature (PRISMA, Cochrane).

package/registry/families/multi-agent-research-lab-ts/files/roles/physics-researcher/methodology/experimental-design.md

@@ -0,0 +1,194 @@
+---
+capability: experimental-design
+role: physics-researcher
+status: active
+source: hand-authored
+retrieved: 2026-04-26
+---
+
+# Experimental Design (Physics)
+
+## Purpose
+
+Help the operator design a physics experiment with hypotheses,
+controls, and an analysis plan. The deliverable is an
+`:::artifact template: experimental-design` the operator can take to
+their advisor / committee / collaborators.
+
+This methodology is for **design**, not execution. The lab does not
+run experiments; it does not give safety advice. Anything that
+crosses into wet-lab safety, IRB, ionising radiation, biosafety, or
+ITAR / EAR triggers `:::escalation`.
+
+## Inputs
+
+- The dispatched sub-question (a "design X" question from the
+  Director).
+- Any prior literature the operator named, plus what
+  `arxiv-literature-search` retrieved in this turn.
+- The operator's stated constraints (instruments available, sample
+  size budget, time horizon).
+
+## Step-by-step procedure
+
+### Step 1 — extract the testable claim
+
+Reframe the operator's question as a **testable claim** with a
+precise observable. Vague: "does the polymer get stiffer near Tg?"
+Testable: "the persistence length of polymer X measured by AFM
+single-molecule pulling increases by > 20% as T → Tg from above."
+
+If you cannot extract a testable claim, return to the Director with
+`finding: question is not yet testable, would need <X> to refine`.
+
+### Step 2 — write H0 / H1 in measurable form
+
+```
+H0: persistence length is constant within measurement noise across
+    the temperature range [T_low, Tg + 20K].
+H1: persistence length increases monotonically with (Tg - T) for
+    T ∈ [T_low, Tg], with effect size > 20% at Tg.
+```
+
+Both hypotheses must reference the actual observable, not a
+qualitative summary. "The polymer gets stiffer" is not an H1.
+
+### Step 3 — variables
+
+Spell out:
+
+- **Independent**: temperature (specify range, step size, ramp
+  protocol).
+- **Dependent**: persistence length (specify estimator — WLC fit?
+  Kratky? — and units).
+- **Controlled**: solvent, polymer concentration, cantilever
+  spring constant, pulling speed, number of pulls per molecule, MW
+  distribution.
+- **Confounded**: anything you can't fully control — solvent
+  evaporation at high T, polymer adsorption to substrate. Name them
+  and propose mitigations.
+
+### Step 4 — statistical plan
+
+Pick the test up-front. Common choices:
+
+- **Two-sample comparison** — t-test or Mann-Whitney U.
+- **Trend across N temperatures** — linear regression with slope CI;
+  if non-parametric, Spearman rank.
+- **Effect-size aware** — Cohen's d or Cliff's delta, not just
+  p-values.
+
+Compute required sample size from the claimed effect size (step 2)
+and a target power (typical 0.8) at α = 0.05. **No experiment ships
+without a power calculation**; an underpowered design wastes the
+operator's time.
+
+For Bayesian designs, specify the prior, the data model, and the
+posterior summary the operator will report.
+
+### Step 5 — controls + replicates
+
+- **Positive control** — a known polymer / system with a published
+  persistence length response.
+- **Negative control** — a system known not to show the effect (a
+  small-molecule analog, or the same polymer at T >> Tg).
+- **Technical replicates** — same molecule pulled N times.
+- **Biological / sample replicates** — N independent molecules.
+
+State both N's. Underspecifying replicates is the most common
+review-killer.
+
+### Step 6 — confounds + mitigations
+
+For each named confound (step 3), propose a mitigation. Examples:
+
+| Confound | Mitigation |
+|---|---|
+| Solvent evaporation at high T | Sealed cell, periodic mass check, abort criteria. |
+| Polymer adsorption | PEG-passivated substrate; check by force-clamp. |
+| Cantilever drift | Periodic recalibration against thermal spectrum. |
+| Operator bias | Blind labels; randomised pull order. |
+
+### Step 7 — output the artifact
+
+```
+:::artifact
+template: experimental-design
+
+Testable claim: [step 1]
+
+Hypotheses:
+  H0: [...]
+  H1: [...]
+
+Variables:
+  Independent: [...]
+  Dependent: [...]
+  Controlled: [...]
+  Confounded (mitigated): [...]
+
+Procedure:
+  1. [...]
+  2. [...]
+
+Statistical plan:
+  Test: [...]
+  α: 0.05
+  Power: 0.80
+  Required N: [from power calc, with formula or tool used]
+  Effect-size estimator: [d / r / etc.]
+
+Replication:
+  Technical: [n]
+  Sample / biological: [n]
+
+Controls:
+  Positive: [...]
+  Negative: [...]
+
+Citations: [for the literature this design draws from]
+:::
+```
+
+### Step 8 — emit handoffs if triggered
+
+If the design depends on a chemistry primitive (specific polymer
+synthesis route, solvent-property data) hand off to chemistry. If it
+depends on a biology primitive (live-cell imaging, protein-specific
+buffer) hand off to biology. The design artifact still ships from
+physics, but the cross-domain context is owned by the right role.
+
+## Mandatory escalation triggers
+
+Stop and emit `:::escalation` if the design involves:
+
+- **Human subjects** — IRB before anything else.
+- **Live-vertebrate animals** — IACUC.
+- **Ionising radiation** — radiation safety officer.
+- **High-pressure / high-voltage / cryogenic** — institutional safety.
+- **BSL-2+** — institutional biosafety committee.
+- **Controlled substances or precursors** — DEA / regulatory affairs.
+- **Classified or export-controlled** — institutional security.
+
+Pair every escalation with the document the operator should bring
+to the professional (protocol draft, hypothesis, power calc, MSDS,
+etc.).
+
+## Anti-patterns
+
+- **Skipping the power calc.** "We'll see what we get" — never
+  ships, gets review-killed.
+- **All-controlled, no-confounded.** Honest design names the
+  confounds it cannot control. Hidden confounds become reviewer
+  ammunition.
+- **Vague observable.** "Stiffer" is not a measurement; persistence
+  length in nm is.
+- **Mixed claims.** Two H1's hidden in one experiment; split them.
+
+## Source
+
+Adapted from Wasserstein & Lazar 2016 (ASA p-value statement),
+Cohen 1988 (power), the methods sections of *Methods* and *Nature
+Protocols*, and Box / Hunter / Hunter *Statistics for Experimenters*.
+Calibrated for a single-experiment design conversation, not a full
+study plan.

package/registry/families/multi-agent-research-lab-ts/manifest.json

@@ -0,0 +1,223 @@
+{
+  "id": "multi-agent-research-lab-ts",
+  "description": "Curated cross-domain research lab \u2014 Lab Director + physics/chemistry/biology researchers with hand-tuned interdisciplinary handoff protocols. Citation-grounded, no diagnosis, no peer-review replacement.",
+  "tags": [
+    "multi-agent",
+    "research",
+    "literature",
+    "cross-domain",
+    "cloudflare-worker"
+  ],
+  "taxonomy": {
+    "language": "markdown",
+    "runtime": "multi-agent",
+    "surface": "multi-agent-team"
+  },
+  "includes": [
+    "agent-base:tangle",
+    "agent-base:secure",
+    "agent-tools:research-corpus",
+    "agent-output:blocks",
+    "agent-base:memory",
+    "agent-base:scheduler",
+    "agent-base:mcp-registry",
+    "agent-channels:telegram",
+    "agent-channels:discord",
+    "agent-channels:slack",
+    "agent-channels:whatsapp",
+    "agent-channels:imessage",
+    "agent-channels:gmail",
+    "agent-channels:linear"
+  ],
+  "defaults": {
+    "projectType": "multi-agent-research-lab",
+    "agentName": "research-lab",
+    "publicVars": [
+      "AGENT_NAME"
+    ],
+    "routes": [
+      {
+        "path": "/api/chat",
+        "auth": "bearer",
+        "callsTools": [
+          "research:paper-search",
+          "research:citation-graph"
+        ]
+      },
+      {
+        "path": "/api/health",
+        "auth": "none",
+        "callsTools": []
+      }
+    ],
+    "declaredCapabilities": [
+      "research-question-formulation",
+      "cross-domain-synthesis",
+      "literature-triage",
+      "arxiv-literature-search",
+      "experimental-design",
+      "reaction-mechanism-analysis",
+      "crossref-literature-search",
+      "protein-structure-analysis",
+      "bioinformatics-survey"
+    ],
+    "runtime": {
+      "sandbox": "tangle",
+      "noNewPrivileges": true,
+      "capDrop": "ALL"
+    },
+    "stakes": "low",
+    "advisoryOnly": true,
+    "outboundDomains": [
+      "api.tangle.tools",
+      "export.arxiv.org",
+      "api.semanticscholar.org",
+      "api.openalex.org",
+      "api.crossref.org",
+      "pubchem.ncbi.nlm.nih.gov",
+      "www.chemspider.com",
+      "rest.uniprot.org",
+      "www.ebi.ac.uk",
+      "eutils.ncbi.nlm.nih.gov"
+    ],
+    "secrets": [
+      "TANGLE_API_KEY"
+    ],
+    "defaultRespondent": "lab-director"
+  },
+  "files": [
+    {
+      "source": "files/agent.json",
+      "target": "agent.json"
+    },
+    {
+      "source": "files/AGENTS.md",
+      "target": "AGENTS.md"
+    },
+    {
+      "source": "files/agents.json",
+      "target": "agents.json"
+    },
+    {
+      "source": "files/README.md",
+      "target": "README.md"
+    },
+    {
+      "source": "files/roles/lab-director/AGENTS.md",
+      "target": "roles/lab-director/AGENTS.md"
+    },
+    {
+      "source": "files/roles/lab-director/methodology/research-question-formulation.md",
+      "target": "roles/lab-director/methodology/research-question-formulation.md"
+    },
+    {
+      "source": "files/roles/lab-director/methodology/cross-domain-synthesis.md",
+      "target": "roles/lab-director/methodology/cross-domain-synthesis.md"
+    },
+    {
+      "source": "files/roles/lab-director/methodology/literature-triage.md",
+      "target": "roles/lab-director/methodology/literature-triage.md"
+    },
+    {
+      "source": "files/roles/physics-researcher/AGENTS.md",
+      "target": "roles/physics-researcher/AGENTS.md"
+    },
+    {
+      "source": "files/roles/physics-researcher/methodology/arxiv-literature-search.md",
+      "target": "roles/physics-researcher/methodology/arxiv-literature-search.md"
+    },
+    {
+      "source": "files/roles/physics-researcher/methodology/experimental-design.md",
+      "target": "roles/physics-researcher/methodology/experimental-design.md"
+    },
+    {
+      "source": "files/roles/chemistry-researcher/AGENTS.md",
+      "target": "roles/chemistry-researcher/AGENTS.md"
+    },
+    {
+      "source": "files/roles/chemistry-researcher/methodology/reaction-mechanism-analysis.md",
+      "target": "roles/chemistry-researcher/methodology/reaction-mechanism-analysis.md"
+    },
+    {
+      "source": "files/roles/chemistry-researcher/methodology/crossref-literature-search.md",
+      "target": "roles/chemistry-researcher/methodology/crossref-literature-search.md"
+    },
+    {
+      "source": "files/roles/biology-researcher/AGENTS.md",
+      "target": "roles/biology-researcher/AGENTS.md"
+    },
+    {
+      "source": "files/roles/biology-researcher/methodology/protein-structure-analysis.md",
+      "target": "roles/biology-researcher/methodology/protein-structure-analysis.md"
+    },
+    {
+      "source": "files/roles/biology-researcher/methodology/bioinformatics-survey.md",
+      "target": "roles/biology-researcher/methodology/bioinformatics-survey.md"
+    }
+  ],
+  "validationChecks": [
+    {
+      "type": "file-exists",
+      "path": "AGENTS.md"
+    },
+    {
+      "type": "file-exists",
+      "path": "agents.json"
+    },
+    {
+      "type": "file-exists",
+      "path": "roles/lab-director/AGENTS.md"
+    },
+    {
+      "type": "agents-md-valid",
+      "path": "roles/lab-director/AGENTS.md"
+    },
+    {
+      "type": "agents-md-valid",
+      "path": "roles/physics-researcher/AGENTS.md"
+    },
+    {
+      "type": "agents-md-valid",
+      "path": "roles/chemistry-researcher/AGENTS.md"
+    },
+    {
+      "type": "agents-md-valid",
+      "path": "roles/biology-researcher/AGENTS.md"
+    }
+  ],
+  "contextHints": {
+    "commands": [],
+    "entrypoints": [
+      "AGENTS.md",
+      "agents.json",
+      "roles/lab-director/AGENTS.md"
+    ],
+    "extensionPoints": [
+      "roles/",
+      "AGENTS.md"
+    ]
+  },
+  "tieredKeywords": {
+    "tier1": [
+      "research-lab-team",
+      "research-lab-agent",
+      "interdisciplinary-research-team",
+      "cross-domain-research-pod"
+    ],
+    "tier2": [
+      "arxiv",
+      "crossref",
+      "openalex",
+      "literature-survey",
+      "interdisciplinary",
+      "cross-domain-citation"
+    ],
+    "tier3": [
+      "agent"
+    ],
+    "archetypes": [
+      "multi-agent-team",
+      "research-pod"
+    ]
+  }
+}