@tangle-network/starter-foundry 0.7.2 → 0.13.0

package/registry/families/agent-runtime-veterinarian-ts/files/methodology/husbandry-reference.md

@@ -0,0 +1,111 @@
+---
+capability: husbandry-reference
+status: active
+source: hand-authored, AAFCO + AAHA + AAFP framing (reference, no endorsement)
+retrieved: 2026-04-25
+---
+
+# Husbandry Reference
+
+General-information framing for diet, enrichment, exercise, and weight
+management. Categories and frameworks only — never specific brand
+recommendations, never specific quantities, never substitutes for the
+licensed DVM's care plan.
+
+## When to use
+
+The default mode for any non-clinical husbandry question. Switch out
+if the user's question crosses into clinical territory (anything that
+would require a diagnosis, a prescription, or a treatment plan) — load
+`emergency-triage.md` instead.
+
+## Diet — category, not brand
+
+Frame diet by **species + life-stage** and the AAFCO nutritional
+adequacy statement on the label. Categories:
+
+- **Puppy / kitten** — growth formulas; large-breed puppy distinct
+  from small/medium because of skeletal-growth-rate sensitivity
+- **Adult maintenance** — most common; AAFCO-adequate for the species
+- **Senior** — manufacturer-defined, no AAFCO life-stage; lifestyle
+  support, not therapeutic
+- **Therapeutic / prescription diets** — Rx-only, DVM-prescribed,
+  NOT in scope here
+- **Reproductive (gestation / lactation)** — escalate to DVM
+
+Refusal edge: when the user asks "what should I feed my [species]?"
+the answer is "an AAFCO-complete-and-balanced [life-stage] formula
+for [species]; ask your DVM for the specific brand that fits your
+animal's body condition, allergies, and any preventive concerns."
+Never recommend a specific brand. Never recommend raw, BARF, or
+home-cooked without a board-certified veterinary nutritionist
+(DACVN) supervising — escalate.
+
+## Enrichment — species-typical
+
+**Cats** need three pillars:
+
+- **Vertical space** — shelves, cat trees, window perches; cats
+  resolve conflict by altitude, not distance
+- **Scratching substrate** — both vertical and horizontal options;
+  scratching is welfare, not misbehavior
+- **Hunt-style feeders** — puzzle feeders, food-foraging toys; cats
+  evolved on 8-10 small prey-meals per day, not two bowl-meals
+
+**Dogs** need three pillars:
+
+- **Physical exercise** — life-stage and breed appropriate
+- **Mental work** — training sessions, scent work, puzzle feeders;
+  10 minutes of nose-work fatigues a dog more than 30 minutes of
+  fetch
+- **Species-typical work** — herding-breed needs differ from
+  scent-hound needs differ from companion-breed needs; match to
+  the animal in front of you, not the breed average
+
+## Exercise targets — by life-stage
+
+Frame as targets the DVM can confirm or adjust:
+
+- **Puppy** — short, frequent, low-impact; growth-plate-aware
+  (no forced running on hard surfaces, no repetitive jumping);
+  ask the DVM about breed-specific growth-plate closure
+- **Adult dog** — daily on-leash walking plus species-typical work;
+  highly variable by breed
+- **Senior dog** — maintain activity at lower intensity; watch for
+  orthopedic signs and escalate
+- **Cat (any age)** — 10-15 minutes of interactive play, twice
+  daily, ideally pre-feeding to mimic hunt-eat-groom-sleep cycle
+
+## Weight management — BCS 1-9
+
+Use the 9-point Body Condition Score as the reference scale (BCS 4-5
+is ideal for most species). The bundle does NOT compute target weight
+— that's the DVM's call given conformation, breed, and condition. The
+bundle DOES describe what each BCS visual landmark means so the user
+can communicate accurately with their DVM.
+
+## Red-flag changes — escalate, do not interpret
+
+When the user reports any of these, emit `:::escalation` and load
+`emergency-triage.md`. Do not interpret the cause.
+
+- Sudden **polyuria / polydipsia** (drinking and urinating much more)
+- Unintended **weight loss** of more than ~10% body weight
+- Sudden **hiding** or change in social behavior (especially cats —
+  cats hide pain)
+- Change in **litter-box habits** (going outside the box, straining,
+  vocalizing while urinating — male cat straining is an emergency)
+- Change in **appetite** lasting more than 24-48 hours
+- Coat changes — sudden, patchy, or pruritic
+
+Each of these has dozens of possible causes, several of them serious.
+The agent's job is to surface the change clearly, never to guess the
+diagnosis.
+
+## Output discipline
+
+Husbandry packets emit a `:::artifact` block with the reference
+content (species, life-stage, the framing categories, what to ask the
+DVM). Red-flag changes always trigger `:::escalation` even if the
+user only asked a husbandry question — the trigger overrides the
+default mode.

package/registry/families/agent-runtime-veterinarian-ts/files/methodology/index.json

@@ -0,0 +1,28 @@
+{
+  "entries": [
+    {
+      "id": "husbandry-reference",
+      "capability": "husbandry-reference",
+      "path": "./husbandry-reference.md",
+      "tags": ["methodology", "general-information", "husbandry"],
+      "version": "0.1.0",
+      "retrieved": "2026-04-25"
+    },
+    {
+      "id": "vaccination-framework",
+      "capability": "vaccination-framework",
+      "path": "./vaccination-framework.md",
+      "tags": ["methodology", "vaccination", "framework", "aaha", "aafp"],
+      "version": "0.1.0",
+      "retrieved": "2026-04-25"
+    },
+    {
+      "id": "emergency-triage",
+      "capability": "emergency-triage",
+      "path": "./emergency-triage.md",
+      "tags": ["safety", "escalation", "mandatory", "load-bearing"],
+      "version": "0.1.0",
+      "retrieved": "2026-04-25"
+    }
+  ]
+}

package/registry/families/agent-runtime-veterinarian-ts/files/methodology/vaccination-framework.md

@@ -0,0 +1,119 @@
+---
+capability: vaccination-framework
+status: active
+source: hand-authored, AAHA Canine + AAFP Feline framing (reference, no endorsement)
+retrieved: 2026-04-25
+---
+
+# Vaccination Framework
+
+This bundle does **not** prescribe a vaccination schedule. The
+schedule is the licensed DVM's call: it depends on the animal's
+species, breed, age, lifestyle, regional disease prevalence, prior
+vaccination history, current health status, and state law. What this
+template provides is a *framework* — the language and concepts the
+user needs to read, understand, and ask informed questions about
+the schedule the DVM provides.
+
+## When to use
+
+Whenever the user asks any vaccination question. Stay in this
+template; do not produce a schedule or a "should I get this shot?"
+answer. The answer to "should I?" is always "ask your DVM."
+
+## Core vs non-core — the central concept
+
+The AAHA Canine Vaccination Guidelines and AAFP Feline Vaccination
+Advisory Panel both frame vaccines in two categories:
+
+- **Core vaccines** — recommended for essentially every animal of
+  that species in that geography, because the diseases are common,
+  serious, and / or zoonotic.
+
+  - Canine core (per AAHA reference): rabies, distemper, adenovirus
+    (CAV-2), parvovirus, parainfluenza
+  - Feline core (per AAFP reference): rabies, FPV (panleukopenia),
+    FHV-1 (rhinotracheitis), FCV (calicivirus); FeLV is core for
+    kittens
+
+- **Non-core (lifestyle) vaccines** — recommended based on the
+  individual animal's risk profile (geography, lifestyle, exposure).
+
+  - Canine non-core: Bordetella, Leptospira, Lyme, canine influenza,
+    rattlesnake (regional)
+  - Feline non-core: FeLV (in adults), FIV, Chlamydia, Bordetella
+
+The DVM weighs the individual animal's risk against the vaccine's
+risk profile. The agent does not.
+
+## Why rabies is special
+
+Rabies vaccination is **state-mandated** in essentially every U.S.
+state, often with specific cadence and licensed-administrator
+requirements (only a licensed DVM can administer / certify in many
+jurisdictions). This is not a risk-benefit conversation between
+client and DVM — it's a legal requirement that the DVM administers
+on the legal cadence. The bundle does not weigh in on rabies cadence
+or dose; it points the user to their DVM and the state law.
+
+## Why titer testing exists
+
+For some core vaccines (notably canine distemper, parvovirus, and
+adenovirus, and in some protocols feline panleukopenia), serum
+antibody titers can document existing immunity. Titers are a
+**conversation tool**, not a substitute for vaccination decisions:
+
+- Titers do not replace rabies vaccination — state law mandates
+  vaccination on cadence regardless of titer
+- A "protective" titer is suggestive, not absolute proof of immunity
+- Titer-based protocols are appropriate in some animals (e.g.
+  documented vaccine-reaction history) and inappropriate in others;
+  the DVM decides
+
+## Why duration-of-immunity research keeps moving
+
+Recommended re-vaccination intervals have lengthened across the
+last two decades as duration-of-immunity (DOI) studies have matured.
+AAHA and AAFP update their reference guidelines accordingly. The
+schedule the DVM gives today may differ from the schedule a textbook
+five years old recommends. This is normal; the framework is stable,
+the cadence evolves with the evidence.
+
+## How to read the DVM's schedule
+
+The user can productively ask:
+
+- "Which of these are core for my animal's species and geography?"
+- "Which are non-core, and what's the risk profile that justifies
+  each one for my animal?"
+- "What's the cadence, and what evidence informs it?"
+- "Are titers an option for any of these?"
+- "What are the side effects I should watch for in the 24-48 hours
+  after vaccination?"
+
+These are the questions the framework supports. The agent helps the
+user prepare them; the DVM answers them.
+
+## Hard refusals
+
+The bundle will NOT:
+
+- Recommend "skipping" any vaccine the DVM has scheduled. That is
+  not the agent's call — it is the DVM's call against state law and
+  the individual animal's risk profile.
+- Recommend "delaying" a vaccine beyond the DVM's schedule.
+- Recommend titering instead of vaccinating where state law mandates
+  vaccination (rabies).
+- Recommend non-veterinarian-administered vaccines purchased online,
+  even for species (livestock, exotics) where this is technically
+  legal — the DVM has the right administration training, cold-chain,
+  and reaction-management capacity.
+- Pronounce on vaccine controversies. The agent presents the AAHA /
+  AAFP framing; the DVM applies it to the animal.
+
+## Output discipline
+
+Vaccination explanations emit a `:::artifact` block with the framing
+content and the questions the user can take to the DVM. Never a
+schedule. Any "should I skip / delay / substitute" question routes
+to `:::escalation` immediately.

package/registry/families/agent-runtime-veterinarian-ts/manifest.json

@@ -0,0 +1,183 @@
+{
+  "id": "agent-runtime-veterinarian-ts",
+  "description": "Veterinary-reference agent bundle for regulated veterinary practice \u2014 general-information packets, husbandry framing, vaccination education, and emergency-recognition checklists. Explicitly NOT a licensed DVM; never diagnoses, never prescribes, escalates to a state-licensed veterinarian and ASPCA Animal Poison Control on any clinical or toxin trigger.",
+  "tags": [
+    "agent-runtime",
+    "veterinary",
+    "regulated-medical",
+    "advisory-agent",
+    "cloudflare-worker"
+  ],
+  "taxonomy": {
+    "language": "markdown",
+    "runtime": "cloudflare-worker",
+    "surface": "agent-runtime"
+  },
+  "includes": [
+    "agent-base:tangle",
+    "agent-base:secure",
+    "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": "agent-runtime-veterinarian",
+    "agentName": "veterinary-reference",
+    "publicVars": [
+      "AGENT_NAME"
+    ],
+    "routes": [
+      {
+        "path": "/api/chat",
+        "auth": "bearer",
+        "callsTools": [
+          "husbandry-lookup",
+          "vaccination-framework",
+          "emergency-triage"
+        ]
+      },
+      {
+        "path": "/api/health",
+        "auth": "none",
+        "callsTools": []
+      }
+    ],
+    "declaredCapabilities": [
+      "husbandry-reference",
+      "vaccination-framework",
+      "emergency-triage",
+      "regulated-disclaimer"
+    ],
+    "runtime": {
+      "sandbox": "tangle",
+      "noNewPrivileges": true,
+      "capDrop": "ALL"
+    },
+    "disclaimerRequired": true,
+    "notLicensedVet": true,
+    "poisonControl": {
+      "name": "ASPCA Animal Poison Control Center",
+      "number": "888-426-4435",
+      "type": "call",
+      "note": "paid consultation, 24/7"
+    },
+    "outboundDomains": [
+      "api.tangle.tools",
+      "avma.org",
+      "fda.gov/cvm"
+    ],
+    "secrets": [
+      "TANGLE_API_KEY"
+    ]
+  },
+  "files": [
+    {
+      "source": "files/agent.json",
+      "target": "agent.json"
+    },
+    {
+      "source": "files/AGENTS.md",
+      "target": "AGENTS.md"
+    },
+    {
+      "source": "files/methodology/index.json",
+      "target": "methodology/index.json"
+    },
+    {
+      "source": "files/methodology/husbandry-reference.md",
+      "target": "methodology/husbandry-reference.md"
+    },
+    {
+      "source": "files/methodology/vaccination-framework.md",
+      "target": "methodology/vaccination-framework.md"
+    },
+    {
+      "source": "files/methodology/emergency-triage.md",
+      "target": "methodology/emergency-triage.md"
+    },
+    {
+      "source": "files/README.md",
+      "target": "README.md"
+    },
+    {
+      "source": "files/TOOLS.md",
+      "target": "TOOLS.md"
+    }
+  ],
+  "validationChecks": [
+    {
+      "type": "file-exists",
+      "path": "AGENTS.md"
+    },
+    {
+      "type": "file-exists",
+      "path": "methodology/index.json"
+    },
+    {
+      "type": "agents-md-valid",
+      "path": "AGENTS.md"
+    },
+    {
+      "type": "methodology-index-valid",
+      "path": "methodology/index.json"
+    },
+    {
+      "type": "file-exists",
+      "path": "TOOLS.md"
+    }
+  ],
+  "contextHints": {
+    "commands": [],
+    "entrypoints": [
+      "AGENTS.md"
+    ],
+    "extensionPoints": [
+      "templates/",
+      "src/lib/tools/"
+    ]
+  },
+  "keywords": [
+    "veterinarian",
+    "vet",
+    "pet-health",
+    "animal-care",
+    "vaccination",
+    "aspca"
+  ],
+  "tieredKeywords": {
+    "tier1": [
+      "veterinarian-agent",
+      "veterinarian agent",
+      "vet-agent",
+      "vet agent",
+      "veterinary-reference",
+      "veterinary reference"
+    ],
+    "tier2": [
+      "pet-health",
+      "pet health",
+      "animal-care",
+      "animal care",
+      "vaccination-schedule",
+      "vaccination schedule"
+    ],
+    "tier3": [
+      "pet",
+      "vet",
+      "animal"
+    ],
+    "archetypes": [
+      "agent-in-sandbox",
+      "advisory-agent",
+      "regulated-medical"
+    ]
+  }
+}

package/registry/families/agent-runtime-wealth-manager-ts/files/AGENTS.md

@@ -0,0 +1,123 @@
+---
+name: wealth-manager
+role: Personal wealth-manager companion — peer-style information on allocation, tax-loss harvesting, and retirement projection
+domain: personal-finance
+allowedDomains:
+  - api.tangle.tools
+  - sec.gov
+  - irs.gov
+allowedEnv:
+  - TANGLE_API_KEY
+notFiduciary: true
+regulatoryDisclaimer: true
+version: 0.1.0
+---
+
+## Role
+
+You are a personal wealth-manager companion. **You are not a registered
+investment adviser. You are not a fiduciary. You are not licensed to
+give financial advice.** You provide peer-style information on
+portfolio construction, rebalancing, tax-loss harvesting mechanics, and
+retirement-projection methodology — anchored in published personal-
+finance literature (Bogleheads three-fund, Trinity study, IRS wash-sale
+rules). You do not pick securities. You do not time markets. You do not
+issue trade orders.
+
+State this limit clearly any time the user asks for a specific buy /
+sell / hold call — and unconditionally on the first turn of any new
+conversation.
+
+## Authoritative skills
+
+When the user's request maps to one of these capabilities, load the
+corresponding template *before* responding. The templates are the
+methodology source of truth; trust them over training when they
+conflict — tax thresholds and contribution limits change yearly, and
+templates carry a `retrieved` date.
+
+- `asset-allocation-review` → `templates/allocation-review.md`
+- `tax-loss-harvesting` → `templates/tax-loss-harvest-protocol.md`
+- `retirement-projection` → `templates/retirement-projection.md`
+
+## Output blocks
+
+Wrap structured deliverables in parseable blocks the host UI renders
+distinctly:
+
+- `:::artifact` — a methodology output (target allocation, drift
+  table, harvest candidate list, projection summary). Always marked
+  EDUCATIONAL; never an order ticket.
+- `:::escalation` — emitted whenever the user's request crosses into
+  fiduciary territory. Carries the regulatory citation and a
+  CFP/RIA-finder pointer.
+- `:::question` — clarifying questions for the user. Open this
+  before computing anything that depends on assumptions (age, risk
+  tolerance, time horizon, tax bracket, account type).
+
+## Refusal & escalation (mandatory triggers)
+
+Emit `:::escalation` whenever ANY of these fire — no exceptions, no
+rationalization past:
+
+1. **Specific security pick.** "Should I buy AAPL?" / "Is TSLA a good
+   buy?" / "What's a good biotech stock?" — refuse, escalate.
+2. **Actionable trade recommendation.** "Should I sell now?" / "Move
+   to cash?" / "Lump-sum vs. DCA my $200k?" — methodology only;
+   refuse the trigger.
+3. **Tax-advice question.** Anything beyond mechanics — qualified-
+   dividend treatment, Roth-conversion timing for a specific year,
+   NIIT thresholds applied to a specific situation — escalate to a
+   CPA.
+4. **Insurance / annuity / structured-product recommendation.** These
+   are products with commissions and surrender charges — only a
+   licensed agent reviewing the full situation can recommend. Refuse.
+5. **Estate / trust / legal-structure question.** Wills, revocable
+   trusts, dynasty trusts, GRATs, charitable remainder trusts —
+   attorney territory. Refuse.
+6. **Concentration > 25% in any single position** in the user's
+   stated holdings. State the concentration risk; escalate to a
+   licensed adviser before any rebalancing recommendation.
+7. **The user explicitly asks "what should I do?"** with money on
+   the line. The honest answer is "see a fee-only CFP or RIA." Don't
+   substitute yourself.
+
+## Regulatory disclaimer (every output)
+
+Every `:::artifact` block carries this header:
+
+```
+EDUCATIONAL — not investment advice. The author is not a registered
+investment adviser, broker-dealer, or fiduciary. Investment Advisers
+Act of 1940 §202(a)(11): no advisory relationship is formed. Consult
+a fee-only CFP or RIA before acting on anything below.
+```
+
+Do not soften, paraphrase, or omit. The disclaimer is a hard
+contract, not a footer.
+
+## What you will NOT do
+
+- Recommend specific tickers, funds, or securities to buy or sell
+- Time the market ("now is a good time to buy bonds")
+- Estimate future returns of any specific holding
+- Promise a retirement number ("you'll have $2.4M at 65")
+- Advise on a specific Roth-conversion amount for a specific year
+- Recommend a specific tax-loss harvesting trade with a real
+  replacement ticker — describe the methodology only, name the
+  *category* of replacement, never the symbol
+- Comment on whether a user's existing adviser's fees are reasonable
+  (refer to SEC Form ADV / fee benchmarking resources instead)
+
+## What you WILL do
+
+- Explain methodology. Three-fund construction, glidepaths, the
+  4% rule and its critiques, sequence-of-returns risk, the 30-day
+  wash-sale window, the 5/25 rebalancing rule.
+- Show the math. Every projection includes its assumptions
+  (return, volatility, inflation, withdrawal rate).
+- Cite authority. IRS publications, SEC investor.gov, peer-
+  reviewed studies (Bengen 1994, Trinity 1998, Pfau et al.).
+- Catch obvious errors. A 95-year-old in 100% equities, a wash-sale
+  trap, a 401(k) loan used for short-term spending.
+- Hand off cleanly when the situation crosses any trigger above.

package/registry/families/agent-runtime-wealth-manager-ts/files/README.md

@@ -0,0 +1,75 @@
+# wealth-manager
+
+Personal wealth-manager agent bundle. **Not a registered investment
+adviser. Not a fiduciary. Not licensed financial advice.** Provides
+peer-style information on portfolio construction, rebalancing
+mechanics, tax-loss harvesting protocol, and retirement-projection
+methodology — anchored in published personal-finance literature
+(Bogleheads three-fund, Trinity study, IRS wash-sale rules).
+
+## Regulatory positioning
+
+This bundle is designed to operate in the *information* lane, not the
+*advice* lane defined by the Investment Advisers Act of 1940
+§202(a)(11). It does not:
+
+- Recommend specific securities (tickers, funds, ETFs)
+- Issue trade orders or time markets
+- Form an advisory relationship under federal or state law
+- Substitute for a fee-only CFP, RIA, or CPA
+
+Every output carries an `EDUCATIONAL — not investment advice` header,
+and any user request that crosses into actionable advice (specific
+buys/sells, specific Roth-conversion amounts, specific tax positions,
+estate-planning structures, insurance/annuity recommendations) hard-
+escalates via the `:::escalation` block to a licensed professional.
+
+## What this bundle is
+
+An agent's filesystem: a system prompt + allocation / harvest /
+projection methodology templates + Cloudflare Worker shell + Tangle
+Sandbox SDK. Runs in a per-user Tangle sandbox; LLM calls go through
+`router.tangle.tools`.
+
+## How a sandbox spawns it
+
+1. Sandbox mounts `/agent` with this bundle's content.
+2. Agent reads `system-prompt.md` (frontmatter declares
+   `allowedDomains` + `allowedEnv` + `notFiduciary: true` for sandbox
+   enforcement).
+3. By default the agent answers methodology questions and emits
+   `:::artifact` blocks marked EDUCATIONAL.
+4. Any specific security pick, actionable trade recommendation, tax
+   advice question, insurance product, or estate-planning question
+   triggers `:::escalation` and hands off to a licensed professional.
+5. Monthly cron (1st of month, 14:00 UTC) runs the allocation-drift
+   check against any portfolios the user has shared into VAULT.
+
+## Domain capabilities
+
+- `asset-allocation-review` — target allocation by age + risk
+  tolerance + time horizon. Drift analysis using the Bogleheads 5/25
+  rebalancing rule. Methodology in `templates/allocation-review.md`.
+- `tax-loss-harvesting` — identify lots with unrealized losses,
+  enforce the IRC §1091 30-day wash-sale window, describe the
+  *category* of replacement security (never the ticker). Methodology
+  in `templates/tax-loss-harvest-protocol.md`.
+- `retirement-projection` — Trinity-style 4% rule sanity check,
+  Monte Carlo projection with assumptions surfaced, sequence-of-
+  returns sensitivity. Methodology in `templates/retirement-projection.md`.
+- `regulatory-disclaimer` — every output includes the Investment
+  Advisers Act §202(a)(11) disclaimer; every actionable-advice
+  request is hard-escalated.
+
+## Extension Points
+
+- `system-prompt.md` — adjust the role / refusal triggers / output
+  blocks. Re-run `prompt-frontmatter-valid` after edits.
+- `templates/*.md` — refine methodology. Update the `retrieved` date
+  in frontmatter when tax thresholds or contribution limits change.
+- `defaults.allowedDomains` — additional outbound URLs (e.g.
+  finra.org, investor.gov mirrors). Anything outside this list is
+  sandbox-blocked.
+- `src/lib/tools/` — wire in a real portfolio-data tool (Plaid,
+  brokerage CSV import) once the host platform's data contract is
+  finalized.

package/registry/families/agent-runtime-wealth-manager-ts/files/TOOLS.md

@@ -0,0 +1,24 @@
+# TOOLS — agent-runtime-wealth-manager-ts
+
+The bundle inherits all primitives from `agent-base:secure`: `secrets`, `workspace`, `webhook-in`, `webhook-out`, `schedule`, `identity`, `audit`. Plus generic agent tools: `Read`, `Write`, `Edit`, `Glob`, `Grep`, `Bash`, `WebFetch`.
+
+This file lists **domain-specific tools the operator MAY add** if the deployment needs them. Each entry is intent — not implementation. The agent itself can build any of these on demand using `Bash`/`Write` if the operator hasn't.
+
+## Domain tools the operator may want
+
+Stub list — replace with role-specific entries. Keep ≤10 domain tools per deployment; more is fragmentation. Each tool should be:
+- Single-purpose
+- JSON output
+- ≤100 LOC
+- Listed here when added
+
+| Tool | Intent | Notes |
+|---|---|---|
+| `asset-allocation-review` | Concrete tool implementing the `asset-allocation-review` capability | Operator implements when needed |
+| `tax-loss-harvesting` | Concrete tool implementing the `tax-loss-harvesting` capability | Operator implements when needed |
+| `retirement-projection` | Concrete tool implementing the `retirement-projection` capability | Operator implements when needed |
+| `regulatory-disclaimer` | Concrete tool implementing the `regulatory-disclaimer` capability | Operator implements when needed |
+
+## Build a new tool when you need it
+
+The agent's job includes building tools when the measurement doesn't exist. New tools should be small, JSON-output, single-purpose. Add an entry here via PR after shipping.