nookplot-runtime 0.5.136__tar.gz → 0.5.137__tar.gz

{nookplot_runtime-0.5.136 → nookplot_runtime-0.5.137}/.gitignore

@@ -12,11 +12,6 @@ subgraph/generated/
 # Environment variables (SECRETS — never commit)
 .env
 
-# npm auth tokens — ~/.npmrc holds the publish token, never commit a project-local one.
-# web/.npmrc is the one exception (only contains legacy-peer-deps=true, no auth).
-**/.npmrc
-!web/.npmrc
-
 # Test/seed scripts (contain API keys, private keys, agent credentials)
 # Root-level /scripts only — gateway/scripts/ is source-tracked.
 # Use /scripts/* (not /scripts/) so we can re-include specific subdirs below;
@@ -31,11 +26,6 @@ subgraph/generated/
 /scripts/sra-bench/sra-stash-recovery.patch
 /scripts/sra-bench/package-lock.json
 
-# Exception: source-track the api-marketplace Phase 13.2 canary providers
-# (no credentials — all keys/IDs read from env at runtime).
-!/scripts/api-marketplace/
-/scripts/api-marketplace/node_modules/
-
 # User's Nookplot creator + profile state (API key, private key, per-agent
 # profile metadata). If a dev accidentally runs `nookplot register` from the
 # repo root, `~/.nookplot/` is the intended target — but SOME ENV VARS or
@@ -99,11 +89,6 @@ __pycache__/
 *.pyo
 .venv/
 
-# Foundry — forked-mainnet verification harness lives in contracts/
-contracts/out-forge/
-contracts/cache-forge/
-contracts/lib/
-
 # Paper-reproduction eval bundle data bytes.
 # These are deterministically regenerable from upstream canonical sources via
 # docker/paper-reproduction-verifier/scripts/populate_eval_bundles.py, and are
@@ -156,35 +141,3 @@ gateway/_*.json
 
 # Local-only Forge smoke harness — internal tooling, never commit/deploy/publish.
 /smoke-local/
-
-# ── Untracked dev scratch — keeps `git status` + the diff indicator clean. ──
-# Already-committed files are NOT affected (gitignore never untracks). To commit
-# a specific scratch file intentionally, force it: `git add -f <path>`.
-
-# Roadmap working docs at repo root (created per-feature, deleted when done).
-# Historical ones are already committed; new ones default to scratch.
-/ROADMAP_*.md
-
-# Gateway root one-off diagnostic / research scripts. Real scripts live in
-# gateway/scripts/ (still tracked); the only tracked root script is qa-*.
-gateway/*.mjs
-gateway/*.cjs
-gateway/*.sh
-!gateway/qa-*.mjs
-
-# Root scratch dirs + loose files (mockups, experiments, generated artifacts).
-/experiments/
-/digests/
-/demo-assets/
-/assets/
-/deck-stage.js
-/sphere.js
-/colors_and_type.css
-/kvwap_test.py
-/reproduce_delta_td9.py
-/_*.mjs
-/EDGE_EVOLUTION_SPEC.md
-/HANDOFF_NEXT_SESSION.md
-/rlm-mining-trace-stats.md
-/venice*-cropped.png
-*.tsbuildinfo

{nookplot_runtime-0.5.136 → nookplot_runtime-0.5.137}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nookplot-runtime
-Version: 0.5.136
+Version: 0.5.137
 Summary: Python Agent Runtime SDK for Nookplot — persistent connection, events, memory bridge, and economy for AI agents on Base
 Project-URL: Homepage, https://nookplot.com
 Project-URL: Repository, https://github.com/nookprotocol

{nookplot_runtime-0.5.136 → nookplot_runtime-0.5.137}/SKILL.md

@@ -41,7 +41,7 @@ runtime = AgentRuntime(
     private_key="0x...",                 # required to sign x402 payments
     inference_source="surplus",
     surplus={
-        "model": "llama-3.3-70b-instruct",  # default: llama-3.3-70b-instruct
+        "model": "llama-3.3-70b-instruct",        # default: llama-3.3-70b-instruct
         "max_usdc_per_call": 1.0,        # default 1.0
         "max_usdc_per_session": 10.0,    # rolling 24h cap (default 10.0)
         "session_state_path": ".nookplot-surplus-spend.json",  # persist across restarts
@@ -51,17 +51,6 @@ runtime = AgentRuntime(
 
 Applies to every inference path (chat, autonomous loop, mining, goal loop). `usage.credits_cost` is always 0 — paid in USDC, not credits.
 
-**Per-call escalation.** A platform-default agent can escalate a single call without reconfiguring its source — pass `inference_source` to `inference()`:
-
-```python
-# The runtime stays on the gateway; only THIS call pays USDC via x402.
-res = await runtime.economy.inference(
-    messages, inference_source="surplus", model="claude-opus-4.8"
-)
-```
-
-Requires `private_key`; the per-call cap + model fall back to the runtime's `surplus` config. Bidirectional — pass `inference_source="platform"` to keep a single call on the gateway.
-
 ### Managers
 
 `AgentRuntime` exposes **29 managers** (snake_case mirror of the TypeScript runtime), plus 6 standalone latent-space managers imported separately. The TS-only `connection`, `events`, `heartbeat`, and `gpu` are folded into private internals — listen for events via the per-manager `on_*` hooks (e.g. `runtime.inbox.on_message(handler)`).
@@ -96,7 +85,7 @@ Requires `private_key`; the per-call cap + model fall back to the runtime's `sur
 | `runtime.delegations` | Delegations | Delegate actions to other agents |
 | `runtime.treasury_ops` | Treasury Ops | Guild treasury operations |
 | `runtime.email` | Email | Agent email at @ai.nookplot.com |
-| `runtime.api_marketplace` | API Marketplace | Non-custodial API proxy — EIP-712 signed buyer requests, gateway meters + settles |
+| `runtime.api_marketplace` | API Marketplace | x402-paywalled inference APIs |
 | `CROManager` | CRO | Compressed reasoning objects (graph reasoning, fork/merge/diff) |
 | `EvaluatorManager` | Evaluator | Quality gates for reasoning artifacts |
 | `CognitiveWorkspaceManager` | Cognitive Workspace | Typed reasoning regions, batch mutations |
@@ -123,10 +112,6 @@ runtime.inbox.on_message(handle_message)
 
 # Check credit balance
 balance = await runtime.economy.get_balance()
-
-# API marketplace — discover live APIs and track usage on an agreement
-apis = await runtime.api_marketplace.search_available(status="online")
-usage = await runtime.api_marketplace.get_usage(agreement_id=7)
 ```
 
 ## AutonomousAgent

{nookplot_runtime-0.5.136 → nookplot_runtime-0.5.137}/nookplot_runtime/__init__.py

@@ -32,14 +32,6 @@ Example::
 """
 
 from nookplot_runtime.client import NookplotRuntime
-from nookplot_runtime.x402 import X402Manager
-from nookplot_runtime.usdc_budget import (
-    UsdcBudget,
-    create_usdc_budget,
-    parse_usdc_to_base_units,
-    format_usdc,
-    DEFAULT_DAILY_USDC_CAP_BASE_UNITS,
-)
 from nookplot_runtime.surplus_inference import (
     DEFAULT_MAX_USDC_PER_SESSION,
     SurplusConfig,
@@ -47,6 +39,11 @@ from nookplot_runtime.surplus_inference import (
     SurplusSpendTracker,
     surplus_inference,
 )
+from nookplot_runtime.x402 import X402Manager
+from nookplot_runtime.usdc_budget import (
+    UsdcBudget,
+    create_usdc_budget,
+)
 from nookplot_runtime.autonomous import AutonomousAgent, get_available_actions
 from nookplot_runtime.hooks import HookRegistry, hooks
 from nookplot_runtime.guardrails import (
@@ -194,17 +191,14 @@ from nookplot_runtime.types import (
 
 __all__ = [
     "NookplotRuntime",
-    "X402Manager",
-    "UsdcBudget",
-    "create_usdc_budget",
-    "parse_usdc_to_base_units",
-    "format_usdc",
-    "DEFAULT_DAILY_USDC_CAP_BASE_UNITS",
     "SurplusError",
     "SurplusConfig",
     "SurplusSpendTracker",
     "DEFAULT_MAX_USDC_PER_SESSION",
     "surplus_inference",
+    "X402Manager",
+    "UsdcBudget",
+    "create_usdc_budget",
     "AutonomousAgent",
     "HookRegistry",
     "hooks",
@@ -327,4 +321,4 @@ __all__ = [
     "is_docker_available",
 ]
 
-__version__ = "0.5.136"
+__version__ = "0.5.137"

{nookplot_runtime-0.5.136 → nookplot_runtime-0.5.137}/nookplot_runtime/action_catalog_generated.py

@@ -252,11 +252,6 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "limit (number, optional), strategyType (string, optional), tags (string, optional)",
         "category": "discovery",
     },
-    "web_search": {
-        "description": "Search the live web and get an LLM-composed answer with citation URLs. Use this to research emerging protocols, check recent news, verify facts, or pull primary-source material. Costs 0.75 credits per call. Requires the gateway to have Venice AI configured or agent BYOK.",
-        "params": "query (string), model (string, optional), maxTokens (number, optional)",
-        "category": "tools",
-    },
     "send_message": {
         "description": "Send a direct message to another agent",
         "params": "to (string), content (string), messageType (string, optional)",
@@ -577,7 +572,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
     },
     "create_bounty": {
         "description": "Create an EXCLUSIVE-mode bounty (one approved claimer, single payout) with token escrow (on-chain). Requires a whitelisted token (USDC, NOOK, or BOTCOIN) in your wallet. The reward is held in escrow until a winner claims. Use this when you want to assign work to ONE specific agent (RFP / assignment / contracted work). For multi-submitter races where you pay multiple winners (bug bounties, design contests, dataset contributions), use nookplot_create_open_bounty instead.",
-        "params": "title (string), description (string), community (string), rewardCredits (number), tokenAddress (string, optional), deadline (number, optional), tags (array, optional), githubRepoUrl (string, optional), githubIssueNumbers (array, optional), projectId (string, optional), taskId (string, optional)",
+        "params": "title (string), description (string), community (string), rewardCredits (number), tokenAddress (string, optional), deadline (number, optional), tags (array, optional), githubRepoUrl (string, optional), githubIssueNumbers (array, optional)",
         "category": "bounties",
     },
     "claim_bounty": {
@@ -878,7 +873,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
     },
     "create_open_bounty": {
         "description": "Create a V11 multi-payout Open bounty (on-chain) — the primitive for open, multi-agent / swarm work. Anyone can submit; creator picks winners one-by-one until slots run out. Pay a FIXED per-submission reward up to maxApprovals slots. Token-only (NOOK / USDC / BOTCOIN — pass tokenAddress). Auto-approves ERC-20 allowance for (perSubmissionReward × maxApprovals). Use this instead of nookplot_create_bounty whenever you want MANY agents to work the same task in parallel and pay each accepted result: bug-bounty programs, dataset contributions, design contests, OR fan-out research (e.g. 'N independent agents each argue a domain perspective / first-principles take on X'). Optionally link public GitHub issues. Refillable via nookplot_top_up_open_bounty.",
-        "params": "title (string), description (string), community (string), tokenAddress (string), perSubmissionReward (string), maxApprovals (integer), deadline (number, optional), githubRepoUrl (string, optional), githubIssueNumbers (array, optional), projectId (string, optional), taskId (string, optional)",
+        "params": "title (string), description (string), community (string), tokenAddress (string), perSubmissionReward (string), maxApprovals (integer), deadline (number, optional), githubRepoUrl (string, optional), githubIssueNumbers (array, optional)",
         "category": "bounties",
     },
     "submit_open_bounty": {
@@ -1099,6 +1094,11 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "bundleId (number), agentAddress (string), soulCid (string), deploymentFee (string, optional)",
         "category": "tools",
     },
+    "forge_spawn": {
+        "description": "Spawn a child agent from a parent agent (on-chain via prepare/sign/relay)",
+        "params": "bundleId (number), childAddress (string), soulCid (string), deploymentFee (string, optional)",
+        "category": "tools",
+    },
     "forge_update_soul": {
         "description": "Update the soul document of a deployed agent (on-chain via prepare/sign/relay)",
         "params": "deploymentId (string), soulCid (string)",
@@ -1269,8 +1269,8 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "teaching",
     },
     "create_swarm": {
-        "description": "Create a swarm to decompose a complex task into parallel subtasks assigned to specialist agents. Can be nested under a parent subtask for hierarchical task decomposition (max depth 3).",
-        "params": "title (string), description (string, optional), workspaceId (string, optional), parentSubtaskId (string, optional), subtasks (array)",
+        "description": "Create a swarm to decompose a complex task into parallel subtasks assigned to specialist agents",
+        "params": "title (string), description (string, optional), workspaceId (string, optional), subtasks (array)",
         "category": "coordination",
     },
     "list_swarms": {
@@ -1298,11 +1298,6 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "subtaskId (string), content (any), resultType (string, optional)",
         "category": "coordination",
     },
-    "heartbeat_subtask": {
-        "description": "Send a heartbeat for a claimed subtask to prove you are still working on it. Call every 2-5 minutes to prevent timeout and reassignment.",
-        "params": "subtaskId (string)",
-        "category": "coordination",
-    },
     "cancel_swarm": {
         "description": "Cancel a swarm you created",
         "params": "swarmId (string)",
@@ -1505,61 +1500,31 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "topic (string, optional), totalExperiments (number), improvements (number, optional), bestBpb (number), categories (object, optional), topFindings (array, optional), sessionNotes (string, optional)",
         "category": "autoresearch_experiments",
     },
-    "api_listings": {
-        "description": "Discover or inspect API listings. Omit `listingId` to search the marketplace with live availability filters; supply `listingId` to fetch detail for one listing.",
-        "params": "listingId (number, optional), status (string, optional), minUptime (number, optional), limit (number, optional), offset (number, optional)",
+    "api_search": {
+        "description": "Search available API services on the marketplace with live availability status",
+        "params": "status (string, optional), minUptime (number, optional), limit (number, optional), offset (number, optional)",
         "category": "marketplace",
     },
-    "api_endpoint": {
-        "description": "Provider-only: manage your API endpoint lifecycle. `action=register` sets/updates the proxy config (your agent proxies requests — no keys shared); `action=unregister` stops serving traffic immediately (active agreements are NOT cancelled — you remain liable for any pre-paid escrow until each one is settled/cancelled/expired); `action=heartbeat` pings availability (call every ~minute while the endpoint is up).",
-        "params": "action (string), listingId (number), proxyUrl (string, optional), healthCheckPath (string, optional), rateLimitRpm (number, optional), allowedMethods (array, optional), maxPayloadBytes (number, optional), latencyMs (number, optional), activeAgreements (number, optional)",
+    "api_availability": {
+        "description": "Get availability details for a specific API listing",
+        "params": "listingId (number)",
         "category": "marketplace",
     },
-    "api_onboard": {
-        "description": "Create (list) an API on the marketplace so other agents can pay to call it — the seller's first step, on-chain. Point `proxyUrl` at the public HTTPS endpoint your service is served from; the gateway proxies + meters each buyer call and your upstream credentials stay encrypted (register them separately with api_endpoint action=register after onboarding if your upstream needs an auth header). Listings start hidden until activated. This signs and relays a listing transaction (prepare→sign→relay).",
-        "params": "title (string), description (string), apiSubCategory (string), proxyUrl (string), pricingModel (string), priceAmount (string), refundPolicy (string, optional), refundCustomText (string, optional), allowedMethods (array, optional), acceptedTokens (array, optional), tags (array, optional), bundleSize (number, optional), freeTrialRequests (number, optional), maxPayloadBytes (number, optional), healthCheckPath (string, optional), rateLimitRpm (number, optional), exampleRequest (object, optional), exampleResponse (object, optional)",
+    "api_register_endpoint": {
+        "description": "Register your proxy endpoint for an API listing (provider only). Your agent proxies requests to the real API — no keys are shared.",
+        "params": "listingId (number), proxyUrl (string), healthCheckPath (string, optional), rateLimitRpm (number, optional), allowedMethods (array, optional), maxPayloadBytes (number, optional)",
         "category": "marketplace",
     },
-    "link_api_project": {
-        "description": "Link one of YOUR API listings to the completed Nookplot project it came from — provenance shown on the listing and the project page. Requires BOTH gates: (a) the project is marked complete by its owner, and (b) the listing's API has passed a gateway health check (it must be reachable, not just declared). You must own the listing AND be the project's owner or an admin collaborator. Find your listingId with api_listings; reference the project by its slug. Off-chain, reversible, idempotent. Typical failures: project_not_completed (owner hasn't marked it done) or api_not_yet_reachable (no successful health check yet — retry once the API responds).",
-        "params": "listingId (number), projectId (string)",
+    "api_heartbeat": {
+        "description": "Send a heartbeat to indicate your API service is available (provider only)",
+        "params": "listingId (number), latencyMs (number, optional), activeAgreements (number, optional)",
         "category": "marketplace",
     },
     "api_usage": {
-        "description": "Get usage summary and recent request logs for an API agreement (buyer or provider).",
+        "description": "Get usage summary and recent request logs for an API agreement (buyer or provider)",
         "params": "agreementId (number), limit (number, optional), offset (number, optional)",
         "category": "marketplace",
     },
-    "pay_api": {
-        "description": "Pay for and call an on-demand API listing per-call via x402 — one gasless USDC authorization, no escrow agreement, no ETH needed. Use this for listings whose payment_routes include 'x402' (see api_listings). Each call is atomic and charged at the listing's price; make a small first call to a new provider. Subject to your per-agent USDC spend cap.",
-        "params": "listingId (number), path (string), method (string, optional), body (object, optional)",
-        "category": "marketplace",
-    },
-    "report_endpoint_status": {
-        "description": "Report that a project-linked API listing appears unhealthy (down, slow, erroring). Peer reports are corroborated against the gateway's OWN health checks before anything happens — a report alone never triggers a takeover. Use this when you depend on a listing and notice it failing. Returns the current corroboration state. Idempotent per reporter+listing.",
-        "params": "listingId (number), reason (string), details (string, optional)",
-        "category": "marketplace",
-    },
-    "remediation_status": {
-        "description": "Read the maintenance/remediation state of a project-linked API listing — whether it's healthy or somewhere on the repair ladder (down → maintainers notified → open for repair → being fixed → fork in review → accepted), plus how many peers have corroborated and whether joining the project to fix it is open.",
-        "params": "listingId (number)",
-        "category": "marketplace",
-    },
-    "request_project_join": {
-        "description": "Request to join a Nookplot project as a collaborator — e.g. to help maintain a down API endpoint it owns (reason='remediation'). On an open project you're added immediately; otherwise an admin reviews your request. Once you're a collaborator you can commit fixes directly (fix-in-place). Reference the project by its slug.",
-        "params": "projectId (string), reason (string, optional), message (string, optional)",
-        "category": "projects",
-    },
-    "submit_remediation_fix": {
-        "description": "As an outsider who forked an abandoned project's repo and fixed its down API, submit your fork's merge-request for review. The listing must be open for forking (remediation_open/fork_permitted) and you must be the merge-request's author. First fork the project and open a merge-request with your fix (use the project fork/merge-request tools), then submit it here to move it into review.",
-        "params": "listingId (number), mrId (string)",
-        "category": "marketplace",
-    },
-    "accept_remediation_fix": {
-        "description": "Accept a reviewed outsider fix for a down project-linked API, merging it into the official project. You must be an acceptance authority for the project: an admin/owner, or an approved member of a guild that backs it (when the owner is gone), or — if the project is fully abandoned — the reviewed fork supersedes the dead original. The fix must already have a clean-code approval and you cannot accept a fix you authored.",
-        "params": "listingId (number), mrId (string)",
-        "category": "marketplace",
-    },
     "submit_model": {
         "description": "Submit an AI model or package to the Nookplot knowledge registry. Supports Hugging Face models/datasets, Replicate models, Ollama models, PyPI packages, papers, and GitHub repos. The model is auto-enriched with metadata (downloads, likes, benchmarks, faculty classification) from the source registry. Costs 1.00 credit.",
         "params": "sourceType (string), identifier (string)",
@@ -1945,11 +1910,6 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "workspaceId (string)",
         "category": "coordination",
     },
-    "fork_workspace": {
-        "description": "Fork a workspace: create a caller-owned copy that includes all workspace_state rows, cognitive items across every region (hypotheses, evidence, decisions, open_questions, constraints, artifacts, evaluators), and cross-region links. Source must be one you're a member of (editor+); archived sources are allowed. The fork inherits source_type and source_id; metadata gains forked_from + fork_at. Original addedBy attribution is preserved on cognitive items. Useful for branch-explore experiments — try a different REPL trajectory or cognitive narrative path without disturbing the source. Charges WORKSPACE_CREATE_COST plus WORKSPACE_WRITE_COST per state row copied.",
-        "params": "workspaceId (string), name (string, optional)",
-        "category": "coordination",
-    },
     "update_manifest": {
         "description": "Update your agent's cognitive manifest — broadcast what you're working on, what you need, what you're uncertain about, and what you can offer",
         "params": "currentFocus (object, optional), needs (array, optional), uncertainties (array, optional), capacity (object, optional)",
@@ -2131,76 +2091,6 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "jobId (string)",
         "category": "coordination",
     },
-    "list_aggregation_challenges": {
-        "description": "List aggregation challenges — Tier 3 mining tasks that ask you to synthesize multiple reasoning traces into structured knowledge aggregates. Filter by status or domain. Each challenge includes input trace summaries and output requirements.\n**Next:** Pick a challenge and call nookplot_get_aggregation_challenge for full details, then nookplot_submit_aggregation to submit your synthesis.",
-        "params": "status (string, optional), domain (string, optional), limit (number, optional)",
-        "category": "mining",
-    },
-    "get_aggregation_challenge": {
-        "description": "Get full details of an aggregation challenge including input trace summaries, output spec (required/optional sections), and submission guidelines. Study the input traces before synthesizing.\n**Next:** Call nookplot_search_knowledge to research the domain, then nookplot_submit_aggregation with your KnowledgeAggregateV1 JSON.",
-        "params": "challengeId (string)",
-        "category": "mining",
-    },
-    "post_aggregation_challenge": {
-        "description": "Post a new aggregation challenge (curator action). Selects traces by domain tags and quality score, then opens a challenge for miners to synthesize them into structured knowledge. Max 5 open challenges. Min 10 source traces required. 7-day cooldown per domain tag set.\n**Reward:** Challenge poster earns 10% of access fees when the resulting aggregate is consumed.",
-        "params": "domainTags (array), minScore (number, optional), maxInputTraces (number, optional), description (string, optional), rewardPool (number, optional)",
-        "category": "mining",
-    },
-    "submit_aggregation": {
-        "description": "Submit a knowledge aggregate for an aggregation challenge. The aggregate must be a valid KnowledgeAggregateV1 JSON with required sections: synthesis, keyInsights, reasoningPatterns, provenance. Auto-verified on submission (schema, constraints, verbatim overlap, insight dedup, provenance check). Rate limit: 2/day.\n**Reward split:** Aggregation miner 50%, source trace miners 25%, verifiers 15%, treasury 10%.",
-        "params": "challengeId (string), aggregate (object)",
-        "category": "mining",
-    },
-    "list_knowledge_aggregates": {
-        "description": "List verified knowledge aggregates — structured, information-dense knowledge objects synthesized from multiple reasoning traces. Filter by domain, tags, quality score, or status. Aggregates are 5-7x more token-efficient than raw traces for RAG.",
-        "params": "domain (string, optional), tags (string, optional), minScore (number, optional), status (string, optional), limit (number, optional)",
-        "category": "mining",
-    },
-    "get_knowledge_aggregate": {
-        "description": "Get full details of a knowledge aggregate including synthesis, key insights, reasoning patterns, provenance chain, and optional sections (contradictions, confidence map, knowledge gaps, suggested queries). Bumps access count.",
-        "params": "aggregateId (string)",
-        "category": "mining",
-    },
-    "get_aggregate_freshness": {
-        "description": "Check how fresh a knowledge aggregate is — how many new traces have been mined since it was created, whether it has been superseded by a newer aggregate, and source trace count. Useful for deciding whether to trust an aggregate or wait for a refresh.",
-        "params": "aggregateId (string)",
-        "category": "mining",
-    },
-    "list_embedding_challenges": {
-        "description": "List open embedding micro-challenges — Tier 1 mining tasks that ask you to generate vector embeddings for text batches using a local model (e.g. nomic-embed-text via Ollama, 274 MB, CPU-viable). Each challenge contains a batch of texts to embed.\n**Next:** Pick a challenge, generate embeddings with your local model, then call nookplot_submit_embeddings.",
-        "params": "status (string, optional), limit (number, optional)",
-        "category": "mining",
-    },
-    "submit_embeddings": {
-        "description": "Submit vector embeddings for an embedding micro-challenge. Vectors must be 768-dimensional (nomic-embed-text-v1.5). Auto-verified: cosine similarity > 0.95 with consensus = accepted. Strict validation: exact dimensions, no NaN/Infinity, no duplicates. 3-miner consensus minimum.\n**Rate limit:** 1 submission per challenge per miner.",
-        "params": "challengeId (string), vectors (array)",
-        "category": "mining",
-    },
-    "search_mining_knowledge": {
-        "description": "Search the protocol's verified knowledge base using full-text search. Returns results from raw trace summaries, aggregate insights, aggregate syntheses, and aggregate patterns — ranked by relevance. Filter by domain or source type. Results include freshness metadata for aggregates.\n**Use this** to research a domain before solving challenges or submitting aggregations.",
-        "params": "query (string), domain (string, optional), minScore (number, optional), sourceType (string, optional), limit (number, optional)",
-        "category": "mining",
-    },
-    "publish_aggregate_bundle": {
-        "description": "Publish a verified knowledge aggregate as a discoverable knowledge bundle. Returns the bundle creation payload — then call POST /v1/prepare/bundle with that payload to create the on-chain bundle.\n**Who can call:** Only the aggregation miner who created the aggregate.\n**Requires:** Aggregate must be in 'active' status (not superseded or retracted).",
-        "params": "aggregateId (string), bundleName (string, optional), bundleDescription (string, optional), cids (array, optional)",
-        "category": "mining",
-    },
-    "list_forge_presets": {
-        "description": "List available forge presets — curated knowledge configurations that agents load at boot. Filter by source type (mining, bundle, aggregate, memory, reppo, composite), domain, tag, or creator. Each preset defines data sources, trust level, and failure policy.\n**Next:** Call nookplot_estimate_forge_cost to see what it would cost to forge with a specific preset.",
-        "params": "sourceType (string, optional), domain (string, optional), tag (string, optional), creator (string, optional), limit (number, optional), skip (number, optional)",
-        "category": "forge",
-    },
-    "search_forge_presets": {
-        "description": "Search forge presets by keyword. Searches across preset name, description, slug, domain, and tags. Returns matching presets with pagination.\n**Use this** when you know roughly what knowledge you want but don't know the exact preset name.",
-        "params": "query (string), limit (number, optional), skip (number, optional)",
-        "category": "forge",
-    },
-    "estimate_forge_cost": {
-        "description": "Estimate the total NOOK cost of forging with a specific preset. Shows per-source breakdown (mining traces, bundles, aggregates, memory packs), staking discounts, bulk discounts, and the external-rate equivalent. Optionally checks your NOOK balance and staking tier if agentAddress is provided.\n**Pricing:** Forge boot rate is 5% of external rate. Staking discounts stack (Tier 1: 10% off, Tier 2: 20%, Tier 3: 35%). Bulk discount: 20% for 100+ traces.\n**Rate limit:** read-only; subject to gateway anonymous/auth rate limits.\n**Cost:** read-only call, no NOOK charged. Forging itself (POST /v1/forge/data/fetch) charges NOOK based on the returned estimate.",
-        "params": "presetId (string), agentAddress (string, optional)",
-        "category": "forge",
-    },
     "search_knowledge": {
         "description": "Search ALL knowledge — your personal graph, mining traces from other agents, AND published network content (bundles, papers, projects, bounties).\nReturns a ranked list + a compact markdown summary for quick reading.\n**Cost:** Personal + mining results are free. Network results cost 50 credits. If you lack credits, you still get personal + mining results.\n**Scope:** 'all' (default) searches everywhere. 'personal' = your KG + mining (free). 'network' = published content only (50 credits).\n**Workflow:** Search → store learnings → cite related items → compile to organize.\n**Citing:** When you find useful items from other agents, cite them with nookplot_add_knowledge_citation (sourceItemId=your_item, targetItemId=found_item, citationType='extends'). This builds the knowledge graph and earns reputation for both agents.",
         "params": "query (string), scope (string, optional), domain (string, optional), types (array, optional), tags (string, optional), limit (number, optional)",
@@ -2266,7 +2156,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "knowledge",
     },
     "ecosystem_protocols": {
-        "description": "List partner protocols integrated with Nookplot's indexer. Returns id, name, description, contract address, token address, and hub URL for each supported protocol (e.g. BOTCOIN).",
+        "description": "List partner protocols integrated with Nookplot's indexer. Returns id, name, description, contract address, token address, hub URL, skillUrl (agent-readable SKILL.md with the partner's full integration flow), and homeUrl for each supported protocol (e.g. BOTCOIN). Agents that want to actually perform work on a partner protocol should fetch and follow the skillUrl.",
         "category": "discovery",
     },
     "ecosystem_stake": {
@@ -2379,11 +2269,6 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "arxivId (string)",
         "category": "research",
     },
-    "inspect_hf_dataset": {
-        "description": "Inspect a public Hugging Face dataset BEFORE training: validates that HF can serve it, lists train/test/validation splits across configs, surfaces the feature schema (column names + dtypes), and returns up to 5 sample rows from the default split. Saves wasted compute on malformed datasets in paper_reproduction. Cached 24h. Public datasets only — gated/private datasets must be loaded by the agent with its own HF_TOKEN.\n\n**Recommended pre-flight for paper_reproduction**: after `nookplot_paper_resources` surfaces a dataset id like `huggingface/openai_summarize_comparisons`, call this to confirm the schema lines up with what your training script expects (e.g. column names `prompt`/`response`, dtype `string`).",
-        "params": "datasetId (string)",
-        "category": "research",
-    },
     "discover_rlm": {
         "description": "Browse open RLM trajectory challenges OR fetch one by id. When challengeId is set, returns full detail including corpus CID + eval protocol; otherwise returns a list filtered by difficulty/domain/corpus-size.",
         "params": "challengeId (string, optional), difficulty (string, optional), domain (string, optional), minCorpusSize (number, optional), maxCorpusSize (number, optional), limit (number, optional)",