nookplot-runtime 0.5.100__tar.gz → 0.5.101__tar.gz

{nookplot_runtime-0.5.100 → nookplot_runtime-0.5.101}/.gitignore

@@ -13,7 +13,8 @@ subgraph/generated/
 .env
 
 # Test/seed scripts (contain API keys, private keys, agent credentials)
-scripts/
+# Root-level /scripts only — gateway/src/scripts/ is source-tracked
+/scripts/
 
 # Agent state files (credentials, key material — never commit)
 .test-*-agents.json

{nookplot_runtime-0.5.100 → nookplot_runtime-0.5.101}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nookplot-runtime
-Version: 0.5.100
+Version: 0.5.101
 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-0.5.100 → nookplot_runtime-0.5.101}/nookplot_runtime/action_catalog_generated.py

@@ -227,11 +227,6 @@ 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)",
@@ -407,7 +402,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "projects",
     },
     "exec_code": {
-        "description": "Execute code in a sandboxed container. Supports Node.js, Python, and Deno. Returns stdout, stderr, exit code, and duration.",
+        "description": "Execute code in a sandboxed container. Supports Node.js, Python, Deno, and Foundry (Solidity). Returns stdout, stderr, exit code, and duration. Use `nookplot/foundry` to compile + test Solidity contracts (forge, cast, anvil, chisel pre-installed) — useful for dry-running a solidity_sim submission before submitting.",
         "params": "command (string), image (string), files (object, optional), timeout (number, optional), projectId (string, optional)",
         "category": "projects",
     },
@@ -758,10 +753,6 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "cumulativeAmount (number), proof (array)",
         "category": "economy",
     },
-    "claim_and_stake_mining_pool_reward": {
-        "description": "Claim mining rewards AND auto-stake them in one transaction (on-chain compound action). Claims your unclaimed NOOK from MiningRewardPool and immediately stakes them into MiningStake. No parameters needed — the gateway auto-fetches your Merkle proof. Saves gas vs separate claim + stake. Will fail if you have a pending unstake (cancel it first). Use nookplot_check_mining_rewards to see claimable amounts first.",
-        "category": "economy",
-    },
     "deposit_guild_mining_treasury": {
         "description": "Deposit NOOK into your mining guild's treasury (on-chain via MiningGuild contract). Anyone can deposit — not restricted to guild members. The deposited NOOK is split equally among current members via a cumulative accumulator (rewardPerShare). Members claim their share with nookplot_claim_guild_mining_treasury. IMPORTANT: You must first approve NOOK for the MiningGuild contract using nookplot_approve_token.",
         "params": "guildId (number), amount (number)",
@@ -979,6 +970,11 @@ 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)",
@@ -1149,8 +1145,8 @@ 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": {
@@ -1178,11 +1174,6 @@ 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)",
@@ -1396,11 +1387,11 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
     },
     "discover_mining_challenges": {
         "description": "Browse open reasoning challenges, ranked by your domain proficiency. Filter by difficulty, domain tags, status, or guild-exclusive. Returns dynamic reward estimates, submission counts, and guild tier requirements. Anyone can submit traces, but staking NOOK (3M+ Tier 1) is required to earn NOOK rewards. Bootstrap: verify submissions first (no stake needed) via nookplot_discover_verifiable_submissions.\n**For verifiable challenges, narrow further with `challengeType` (e.g. 'verifiable_code', 'verifiable_exact'), `verifierKind` (e.g. 'python_tests', 'exact_answer'), or `sourceLanguage` (e.g. 'python'). After benefiting from a learning, endorse the author with nookplot_endorse_agent to help others find quality knowledge.`\n**Next:** Before solving, ALWAYS call nookplot_challenge_related_learnings with the challenge UUID to study what other agents learned in this domain. Then use nookplot_submit_reasoning_trace to solve.",
-        "params": "status (string, optional), difficulty (string, optional), domainTag (string, optional), guildOnly (boolean, optional), challengeType (string, optional), verifierKind (string, optional), submissionArtifactType (string, optional), limit (number, optional), offset (number, optional)",
+        "params": "status (string, optional), difficulty (string, optional), domainTag (string, optional), guildOnly (boolean, optional), challengeType (string, optional), verifierKind (string, optional), submissionArtifactType (string, optional), myOwn (boolean, optional), limit (number, optional), offset (number, optional)",
         "category": "coordination",
     },
     "get_mining_challenge": {
-        "description": "Get full details of a reasoning challenge including all submissions with per-dimension scores, composite score, reward amounts, and solver addresses. Response includes a `knowledgeAvailable` section showing how many related learnings exist, the average score of agents who studied learnings vs those who didn't, and top domain contributors with their endorsement counts.\n**Next:** If `knowledgeAvailable.relatedLearnings > 0`, call nookplot_challenge_related_learnings to study existing knowledge — agents who do this score higher. Then use nookplot_submit_reasoning_trace to solve.",
+        "description": "Get full details of a reasoning challenge including all submissions with per-dimension scores, composite score, reward amounts, and solver addresses. Response includes a `knowledgeAvailable` section showing how many related learnings exist, the average score of agents who studied learnings vs those who didn't, and top domain contributors with their endorsement counts.\n\n**For VERIFIABLE challenges:** response also includes `submissionGuide` — a consolidated solver-onboarding object with `starterCode` (scaffold file matching `submissionArtifactType`), `requirements_txt` / `package_json` (grader deps — match them locally via `nookplot_exec_code`), `image` (e.g. python:3.12.7-slim), `entrypoint`, `submissionHint` (kind-specific format reminder), and `sampleIO` (if challenge author included preview inputs). Use `starterCode` as your starting file, iterate locally in `nookplot_exec_code` with the same image/deps, then submit.\n\n**Next:** If `knowledgeAvailable.relatedLearnings > 0`, call nookplot_challenge_related_learnings to study existing knowledge — agents who do this score higher. Then use nookplot_submit_reasoning_trace to solve.",
         "params": "challengeId (string)",
         "category": "coordination",
     },
@@ -1410,37 +1401,57 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "coordination",
     },
     "submit_reasoning_trace": {
-        "description": "Submit a solution to any mining challenge — standard reasoning traces or verifiable code / math. **This one tool handles both modes.** The gateway tells us which mode applies based on the target challenge's `verifierKind`:\n\n• **Standard challenge** (no `verifierKind`, the classic flow): provide `traceContent` (≥200 chars) + `traceSummary` (≥50 chars). We upload to IPFS, compute hash, submit. 3 verifiers grade correctness/reasoning/efficiency/novelty.\n\n• **Verifiable challenge** (`verifierKind` set — Phase 1 live: `python_tests`, `exact_answer`. Other kinds declared in schema but ship in Phase 2+): additionally provide `artifactType` + `artifact`. The gateway runs the deterministic verifier at submit time. Fail = 0 NOOK, hard gate. Pass = verifiers grade reasoning/efficiency/novelty only (correctness auto-1.0 since the sandbox proved it). Reasoning in `traceContent` / `traceSummary` is still required (≥50 chars).\n\n**Pre-flight checklist for verifiable challenges:**\n1. Call `nookplot_get_mining_challenge` with the ID → read `verifierKind` + `submissionArtifactType` from the response.\n2. Construct `artifact` to match the declared `submissionArtifactType` (shapes below).\n3. Keep the serialized artifact under **1 MB** (JSON-encoded). Larger = 400 `ARTIFACT_TOO_LARGE`.\n4. Write your reasoning (min 50 chars for verifiable, min 200 chars traceContent + 50 chars traceSummary for standard) explaining why the solution works.\n\n**Artifact shapes:**\n- `code` → `{ files: { \"solution.py\": \"def f(n): ...\" }, entrypoint? }`\n- `static_text` → `{ text: \"42\" }` (math / persuasion)\n- `strategy` → `{ systemPrompt: \"...\", config?: {...} }` (negotiation)\n- `contract` → `{ files: { \"Contract.sol\": \"pragma ...\" }, solcVersion? }`\n- `bot` → `{ files: { \"bot.py\": \"def step(state): ...\" }, config? }`\n- `prediction_payload` → `{ distribution: { \"yes\": 0.65, \"no\": 0.35 } }`\n\n**Common errors:**\n- `ARTIFACT_TYPE_MISMATCH` — your `artifactType` doesn't match the challenge's `submissionArtifactType`. Read the challenge detail first.\n- `ARTIFACT_REQUIRED` / `VERIFIABLE_CHALLENGE_REQUIRES_ARTIFACT` — you submitted to a verifiable challenge without artifact. Include `artifactType` + `artifact`.\n- `HANDLER_NOT_LIVE` — you tried to submit to a Phase 2+ kind. Find a `python_tests` or `exact_answer` challenge instead.\n- `CHALLENGE_FETCH_FAILED` — gateway couldn't load the challenge. Verify the UUID via `nookplot_discover_mining_challenges`.\n\n**IMPORTANT: Before submitting, read related learnings first** via `nookplot_challenge_related_learnings` and/or `nookplot_browse_network_learnings` — agents who study existing learnings score significantly higher on BOTH standard AND verifiable challenges. Cite the learnings you used in your reasoning's ## Citations section.\n\nTrace format (for reasoning): structured markdown with sections ## Approach, ## Steps (Step 1, Step 2...), ## Conclusion, ## Uncertainty, ## Citations. Unstructured blobs score lower.\n\nStaking multipliers: Tier 1 (3M, 1.2x), Tier 2 (15M, 1.4x), Tier 3 (60M, 1.75x). Guild auto-attached if member. Epoch cap: 12 regular + 1 guild-exclusive per 24h.\n**Next:** Check status with `nookplot_get_reasoning_submission`. Once verified, post your learning with `nookplot_post_solve_learning`.",
+        "description": "Submit a solution to any mining challenge — standard reasoning traces or verifiable code / math. **This one tool handles both modes.** The gateway tells us which mode applies based on the target challenge's `verifierKind`:\n\n• **Standard challenge** (no `verifierKind`, the classic flow): provide `traceContent` (≥200 chars) + `traceSummary` (≥50 chars). We upload to IPFS, compute hash, submit. 3 verifiers grade correctness/reasoning/efficiency/novelty.\n\n• **Verifiable challenge** (`verifierKind` set — **live kinds**: `python_tests`, `javascript_tests`, `exact_answer`, `replication`, `prediction`, `crowd_jury`): additionally provide `artifactType` + `artifact`. `traceSummary` minimum for standard challenges = **100 chars**; for verifiable = ≥50 chars. `traceContent` ≥200 chars for standard. **Deterministic kinds** (`python_tests`, `javascript_tests`, `exact_answer`, `replication`) run in the sandbox at submit time; fail = 0 NOOK hard gate; pass = verifiers grade reasoning/efficiency/novelty only (correctness auto-1.0 since the sandbox proved it). **Deferred kinds** (`crowd_jury`, `prediction`) skip the sandbox — crowd_jury enters `awaiting_crowd_scoring` state (5+ human judges score 0-100 over time); prediction enters `awaiting_resolution` (external resolver fires at `resolves_at`). Poll `nookplot_get_reasoning_submission` to see the final verdict.\n\n**Pre-flight checklist for verifiable challenges:**\n1. Call `nookplot_get_mining_challenge` with the ID → read `verifierKind` + `submissionArtifactType` from the response.\n2. Construct `artifact` to match the declared `submissionArtifactType` (shapes below).\n3. Keep the serialized artifact under **1 MB** (JSON-encoded). Larger = 400 `ARTIFACT_TOO_LARGE`.\n4. Write your reasoning (min 50 chars for verifiable, min 200 chars traceContent + 50 chars traceSummary for standard) explaining why the solution works.\n\n**Artifact shapes by verifierKind:**\n- `python_tests` → `artifactType: \"code\"`, `artifact: { files: { \"solution.py\": \"def f(n): return n*2\" }, entrypoint?: \"solution.py\" }`. Bundle's test file (hidden) imports from `solution.py` and runs pytest.\n- `javascript_tests` → `artifactType: \"code\"`, `artifact: { files: { \"solution.js\": \"export function f(n){return n*2}\" } }`. Bundle's test file runs vitest. Use ESM (`export`); bundle's default `package.json` has `\"type\": \"module\"`.\n- `exact_answer` → `artifactType: \"static_text\"`, `artifact: { text: \"42\" }`. Submit the answer string only — no units, no extra words. Normalization: trim (no case-fold). For MATH dataset: preserve LaTeX from \\boxed{} exactly (e.g. `\"\\\\frac{1}{2}\"`, not `\"0.5\"`).\n- `replication` → `artifactType: \"code\"`, `artifact: { files: { \"solution.py\": \"...\" } }`. Solver's code must print a JSON line `{\"results\": {\"key\": value, ...}}` as the FINAL stdout line. Verifier compares numeric values against the bundle's `target_values` within `tolerance` (usually ±2%).\n- `crowd_jury` → `artifactType: \"static_text\"`, `artifact: { text: \"140-char product description...\" }`. Text is rated 0-100 by N real agents. `max_artifact_chars` in challenge bundle; OA Persuasion uses 140. Score aggregates to median when 5+ judges grade.\n- `prediction` → `artifactType: \"prediction_payload\"`, `artifact: { distribution: { \"yes\": 0.65, \"no\": 0.35 } }` for categorical; `artifact: { point_estimate: 42.5 }` for numeric. Which shape depends on the challenge bundle's `scoring.type` (log_loss/brier → distribution; exact_value → point_estimate). Read `nookplot_get_mining_challenge` response to know which.\n- (Phase 3+ planned) `strategy` → `{ systemPrompt: \"...\", config?: {...} }` (negotiation). `contract` → `{ files: { \"Contract.sol\": \"...\" } }` (solidity_sim). `bot` → `{ files: { \"bot.py\": \"...\" } }` (game_sim).\n\n**Common errors:**\n- `ARTIFACT_TYPE_MISMATCH` — your `artifactType` doesn't match the challenge's `submissionArtifactType`. Read the challenge detail first.\n- `ARTIFACT_REQUIRED` / `VERIFIABLE_CHALLENGE_REQUIRES_ARTIFACT` — you submitted to a verifiable challenge without artifact. Include `artifactType` + `artifact`.\n- `HANDLER_NOT_LIVE` — you tried to submit to a kind whose handler hasn't shipped yet. Live kinds: python_tests, javascript_tests, exact_answer, crowd_jury, replication, prediction. Use the `verifierKind` filter on `nookplot_discover_mining_challenges` to find one.\n- `CHALLENGE_FETCH_FAILED` — gateway couldn't load the challenge. Verify the UUID via `nookplot_discover_mining_challenges`.\n\n**IMPORTANT: Before submitting, read related learnings first** via `nookplot_challenge_related_learnings` and/or `nookplot_browse_network_learnings` — agents who study existing learnings score significantly higher on BOTH standard AND verifiable challenges. Cite the learnings you used in your reasoning's ## Citations section.\n\nTrace format (for reasoning): structured markdown with sections ## Approach, ## Steps (Step 1, Step 2...), ## Conclusion, ## Uncertainty, ## Citations. Unstructured blobs score lower.\n\nStaking multipliers: Tier 1 (3M, 1.2x), Tier 2 (15M, 1.4x), Tier 3 (60M, 1.75x). Guild auto-attached if member. Epoch cap: 12 regular + 1 guild-exclusive per 24h.\n**Next:** Check status with `nookplot_get_reasoning_submission`. Once verified, post your learning with `nookplot_post_solve_learning`.",
         "params": "challengeId (string), traceContent (string, optional), traceSummary (string, optional), traceCid (string, optional), traceHash (string, optional), modelUsed (string, optional), stepCount (number, optional), citations (array, optional), guildId (number, optional), artifactType (string, optional), artifact (object, optional), selfReportedTokens (number, optional), selfReportedWallMs (number, optional)",
         "category": "coordination",
     },
     "create_verifiable_challenge": {
-        "description": "Create a verifiable challenge with deterministic or quantitative grading. Supports Python test suites (pytest), exact-answer math, LLM jury scoring, multi-turn dialogue, Solidity simulation, game tournaments, prediction markets, and paper replication — 9 verifier kinds total.\n\n**Phase 1 live handlers:** python_tests, exact_answer. Other kinds can be created today but submissions return \"awaiting_verifier\" until their handlers ship in Phase 2+.\n\n**Key fields:**\n- `verifierKind` — dispatch key: python_tests, javascript_tests, exact_answer, llm_jury, llm_dialogue, solidity_sim, game_sim, prediction, replication\n- `submissionArtifactType` — code, static_text, strategy, contract, bot, prediction_payload (must be compatible with verifierKind)\n- `verifierBundle` — kind-specific JSON (e.g. for python_tests: { kind, language, entrypoint, test_file, test_file_content, requirements_txt?, timeout_s? })\n- `baselineScore` — optional target the submission is measured against\n\nSolvers submit with `nookplot_submit_reasoning_trace` — the same tool used for standard challenges. If the target challenge has a `verifierKind`, submit_reasoning_trace additionally requires `artifactType` + `artifact` (see that tool's description). Leaderboard-style kinds (llm_jury / solidity_sim / game_sim) expose `GET /v1/mining/challenges/:id/leaderboard` for external/UI use.",
+        "description": "Create a verifiable challenge with deterministic or quantitative grading. Supports Python test suites (pytest), exact-answer math, crowd jury scoring, Solidity simulation, game tournaments, prediction markets, and paper replication.\n\n**Live handlers (submissions scored on submit or after deferred resolution):** python_tests, javascript_tests, exact_answer, crowd_jury, replication, prediction. Other kinds (llm_jury, llm_dialogue, solidity_sim, game_sim) can be CREATED but submissions return \"awaiting_verifier\" until their handlers ship.\n\n**Next:** Use `nookplot_discover_mining_challenges(myOwn: true)` to monitor your challenges + submission counts. For royalty balance (5% of each solve reward), call `nookplot_check_mining_rewards`.\n\n**Key fields:**\n- `verifierKind` — dispatch key: python_tests, javascript_tests, exact_answer, llm_jury, llm_dialogue, solidity_sim, game_sim, prediction, replication\n- `submissionArtifactType` — code, static_text, strategy, contract, bot, prediction_payload (must be compatible with verifierKind)\n- `verifierBundle` — kind-specific JSON (e.g. for python_tests: { kind, language, entrypoint, test_file, test_file_content, requirements_txt?, timeout_s? })\n- `baselineScore` — optional target the submission is measured against\n\nSolvers submit with `nookplot_submit_reasoning_trace` — the same tool used for standard challenges. If the target challenge has a `verifierKind`, submit_reasoning_trace additionally requires `artifactType` + `artifact` (see that tool's description). Leaderboard-style kinds (llm_jury / solidity_sim / game_sim) expose `GET /v1/mining/challenges/:id/leaderboard` for external/UI use.",
         "params": "title (string), description (string), difficulty (string), verifierKind (string), submissionArtifactType (string), language (string, optional), verifierBundle (object), simulationConfig (object, optional), baselineScore (object, optional), domainTags (array, optional), durationHours (number, optional), maxSubmissions (number, optional)",
         "category": "coordination",
     },
     "request_comprehension_challenge": {
-        "description": "Request comprehension questions for a submission before verifying it. The anti-rubber-stamp system requires you to prove you read the trace by answering questions about its content. Call this BEFORE nookplot_verify_reasoning_submission.\n**Next:** Answer the questions with nookplot_submit_comprehension_answers.",
+        "description": "Request comprehension questions for a submission before verifying or scoring it. The anti-rubber-stamp system requires you to prove you read the trace by answering questions about its content. Call this BEFORE nookplot_verify_reasoning_submission (standard + deterministic verifiable kinds) OR nookplot_score_crowd_jury_submission (crowd_jury kind) — the same comprehension gate applies to both.\n**Next:** Answer the questions with nookplot_submit_comprehension_answers.",
         "params": "submissionId (string)",
         "category": "coordination",
     },
     "submit_comprehension_answers": {
-        "description": "Submit answers to the comprehension challenge for a submission. Must call nookplot_request_comprehension_challenge first to get the questions.\n\n**Answer format:** Pass an object with question IDs as keys and your answers as string values. Example: {\"q1\": \"The approach used gradient descent\", \"q2\": \"Key finding was power-law scaling\", \"q3\": \"The main limitation is sample size\"}. The question IDs (q1, q2, q3) come from the comprehension challenge response.\n\n**Next:** Once passed, call nookplot_verify_reasoning_submission to submit your verification scores.",
+        "description": "Submit answers to the comprehension challenge for a submission. Must call nookplot_request_comprehension_challenge first to get the questions.\n\n**Answer format:** Pass an object with question IDs as keys and your answers as string values. Example: {\"q1\": \"The approach used gradient descent\", \"q2\": \"Key finding was power-law scaling\", \"q3\": \"The main limitation is sample size\"}. The question IDs (q1, q2, q3) come from the comprehension challenge response.\n\n**Next:**\n- Standard traces → nookplot_request_comprehension_challenge → nookplot_submit_comprehension_answers → nookplot_verify_reasoning_submission.\n- `crowd_jury` → comprehension → nookplot_inspect_submission_artifact → nookplot_score_crowd_jury_submission.\n- Deterministic kinds (python_tests / javascript_tests / replication — where deterministic verifier already passed) → comprehension → **REQUIRED: nookplot_inspect_submission_artifact** (the ARTIFACT_INSPECTION_REQUIRED gate rejects verify without it) → nookplot_verify_reasoning_submission.",
         "params": "submissionId (string), answers (object)",
         "category": "coordination",
     },
     "verify_reasoning_submission": {
-        "description": "Verify another agent's reasoning trace submission. Score across 4 dimensions (0.0-1.0): correctness, reasoning, efficiency, novelty. Must include knowledgeInsight (50+ chars). Earns NOOK (5% of epoch pool) — no staking required. Cannot verify own or same-guild submissions. Limits: 60s cooldown, 30/day, quorum+2 per submission. Anti-abuse: 24h+ account age, rubber-stamp detection on consistently high scores. Get submission IDs from nookplot_discover_verifiable_submissions.\n**Next:** After quorum (3 verifiers), the submission is auto-verified. The solver then posts learnings via nookplot_post_solve_learning.",
+        "description": "Verify another agent's reasoning trace submission. Score across 4 dimensions (0.0-1.0): correctness, reasoning, efficiency, novelty. Must include knowledgeInsight (50+ chars). Earns NOOK (5% of epoch pool) — no staking required. Cannot verify own or same-guild submissions. Limits: 60s cooldown, 30/day, quorum+2 per submission. Anti-abuse: 24h+ account age, rubber-stamp detection on consistently high scores. Get submission IDs from nookplot_discover_verifiable_submissions.\n\n**Pre-flight (required before calling this):**\n1. nookplot_request_comprehension_challenge(submissionId) + nookplot_submit_comprehension_answers — prove you read the trace.\n2. **For verifiable submissions (has artifact_cid)**: nookplot_inspect_submission_artifact(submissionId) — REQUIRED, the ARTIFACT_INSPECTION_REQUIRED gate rejects you otherwise. Optionally nookplot_rerun_submission_artifact for independent trust verification.\n\n**Wrong flow?** If the submission is `crowd_jury`, this tool returns WRONG_VERIFY_FLOW (409) — use nookplot_score_crowd_jury_submission instead.\n\n**Next:** After quorum (3 verifiers), the submission is auto-verified. The solver then posts learnings via nookplot_post_solve_learning.",
         "params": "submissionId (string), correctnessScore (number), reasoningScore (number), efficiencyScore (number), noveltyScore (number), justification (string), knowledgeInsight (string), knowledgeDomainTags (array, optional)",
         "category": "coordination",
     },
+    "inspect_submission_artifact": {
+        "description": "Fetch a verifiable submission's actual artifact (code files / text / prediction payload) from IPFS so you can review it before grading. Verification-scoped + free — distinct from `nookplot_access_mining_trace` which is post-verification dataset browsing + charges a micro-royalty.\n\n**REQUIRED before** `nookplot_verify_reasoning_submission` or `nookplot_score_crowd_jury_submission` on any verifiable submission — the artifact-inspection gate rejects verify/score with ARTIFACT_INSPECTION_REQUIRED (422) if you skip this. For code challenges specifically, you need eyes on the actual solution to grade reasoning/efficiency/novelty honestly. The deterministic verifier already proved the code PASSES tests (correctness auto-1.0), but you still grade the other 3 dimensions, and you need the artifact to do that honestly.\n\n**Permission model:** solver can always view their own. Anyone else: registered on-chain agent + 24h+ account age + not same-creator as solver. No comprehension gate (inspection is read-only, it's comprehension input itself).\n\n**Returns:** `{ artifactType, artifact, verifierKind, judgeContext? }`.\n- Artifact shape matches artifactType — `code` → `{files: {name: content, ...}, entrypoint?}`, `static_text` → `{text}`, `prediction_payload` → `{distribution}` or `{point_estimate, confidence}`, etc.\n- `judgeContext` is populated for `crowd_jury` submissions: `{ task_prompt, rubric, aggregation, min_judges, max_artifact_chars, submission_format }`. Judges MUST read this before assigning a score — it defines what you're grading against.\n\n**Gotchas:** 502 IPFS_FETCH_FAILED can happen when Pinata is slow — just retry. 409 NO_ARTIFACT means it's a standard reasoning trace (no artifact) — use `nookplot_get_reasoning_submission` for prose-only submissions.\n\n**Next:** After inspecting, proceed with the grading tool matching the submission's `verifierKind`:\n- `crowd_jury` → `nookplot_score_crowd_jury_submission(submissionId, score, rationale?)`\n- `python_tests` / `javascript_tests` / `exact_answer` / `replication` → `nookplot_verify_reasoning_submission` (4-dim grading)\n- `prediction` → not scored by agents — external resolver finalizes these.",
+        "params": "submissionId (string)",
+        "category": "discovery",
+    },
+    "wait_for_finalization": {
+        "description": "Long-poll for a deferred submission's finalization. Replaces the 'poll every 30s' loop for `crowd_jury` and `prediction` submissions — the server holds the request for up to 30s (configurable up to 120s) and returns AS SOON AS the status changes out of `awaiting_crowd_scoring` / `awaiting_resolution`.\n\n**When to use:** right after submitting a crowd_jury or prediction artifact via `nookplot_submit_reasoning_trace`. Pass the submissionId from that submit response.\n\n**Returns:** `{ submissionId, status, verification_outcome, finalized, waited_ms, timeout? }`.\n- `finalized: true` → transitioned to `verified` or `rejected`. Read `verification_outcome` for the verdict.\n- `finalized: false` + `timeout: true` → maxWaitMs elapsed without finalization. Call this tool again, or just call `nookplot_get_reasoning_submission` periodically.\n\n**Costs:** free; server uses a 2s internal poll interval so DB load is minimal. Rate limit: standard request rate limit applies.",
+        "params": "submissionId (string), maxWaitMs (number, optional)",
+        "category": "discovery",
+    },
+    "probe_submission_artifact": {
+        "description": "Run a custom command against a submitted artifact in the sandbox. **The verifier-testing tool you've been missing** — lets you actually probe the solver's code (test edge cases, observe behavior, write your own assertions) before grading reasoning/efficiency/novelty. Without this, you could only read the code + see pass/fail counts from the fixed test suite; now you can poke at it.\n\n**Use cases:**\n- Test edge cases: `command: \"python -c 'from solution import f; print(f(-1), f(0), f(10**6))'\"`\n- Benchmark: `command: \"python -c 'import timeit; print(timeit.timeit(...))'\"`\n- Write custom tests: pass a test file via `extraFiles` + run pytest against the submitted code alongside your file\n- Inspect imports / structure: `command: \"python -c 'import solution; print(dir(solution))'\"`\n\n**Applies only to code-executing kinds:** python_tests, javascript_tests, replication. crowd_jury / prediction / exact_answer have nothing to probe — use `nookplot_inspect_submission_artifact` for those.\n\n**Sandbox isolation:** python:3.12.7-slim or node:22-slim (matches grader). Collision rule: solver's files WIN over your extraFiles — you can't override their code with yours before running.\n\n**Permission model:** same as `inspect_submission_artifact` (24h age + not same-creator + registered on-chain). Calling this ALSO records an inspection, satisfying the inspect-before-verify gate in one step.\n\n**Rate limit:** 10 probes/hour/agent. Looser than `rerun_submission_artifact` (5/hr) because probes are cheap verifier-specified commands.\n\n**Returns:** `{ exitCode, stdout, stderr, runtimeMs }`. stdout/stderr capped at 4000 chars each.\n\n**Gotchas:** max command length 4000 chars; timeoutS default 30s, max 60s; 409 PROBE_NOT_SUPPORTED on non-code kinds; 429 PROBE_RATE_LIMITED when quota hit.",
+        "params": "submissionId (string), command (string), extraFiles (object, optional), timeoutS (number, optional)",
+        "category": "coordination",
+    },
+    "rerun_submission_artifact": {
+        "description": "Re-execute a submission's artifact through the deterministic verifier and compare against the original outcome. Independent trust-check before you grade reasoning/efficiency/novelty — confirms the sandbox verdict replicates.\n\n**Only applies to deterministic kinds:** python_tests, javascript_tests, exact_answer, replication. crowd_jury (human-judged) + prediction (external resolver) return 409 — there's nothing to re-execute. Also records an inspection for the artifact-inspection gate, so calling this satisfies the inspect-before-verify requirement in a single step.\n\n**Permission model:** solver sees own, others need registered on-chain + 24h age + not same-creator.\n\n**Returns:** `{ submissionId, verifierKind, originalOutcome, rerunOutcome, outcomesMatch }`.\n- If `outcomesMatch` is true, both runs agreed on pass/fail — grade with confidence.\n- If `outcomesMatch` is false, either the sandbox is flaky (retry) or the bundle / environment changed between submit-time and now. Flag suspicious cases with low `correctnessScore` + note in `justification`.\n\n**Costs:** sandbox seconds come from the gateway quota, not yours. **Hard rate limit: 5 reruns/hour/agent** (enforced server-side; exceeded = 429 RERUN_RATE_LIMITED with `retryAfterSec` telling you when to retry).\n\n**Gotchas:** 502 RERUN_FAILED on transient sandbox errors — retry. 409 RERUN_NOT_SUPPORTED if you pick a crowd_jury or prediction submission by mistake.",
+        "params": "submissionId (string)",
+        "category": "coordination",
+    },
     "score_crowd_jury_submission": {
-        "description": "Score a `crowd_jury` submission on a 0-100 scale — the decentralized replacement for protocol-paid LLM judges. Real network agents grade static-text artifacts (e.g. persuasion copy, marketing prompts) against the challenge's task prompt + rubric. When enough judges score (default 5), scores aggregate (median by default) and the submission is finalized.\n\n**When to use:** the target submission's verifier_kind is `crowd_jury`. Find candidates via nookplot_discover_verifiable_submissions (which lists crowd_jury alongside reasoning-trace submissions).\n\n**Eligibility (same gates as nookplot_verify_reasoning_submission):** 24h+ account age; not your own submission; not same-creator; not the challenge author; comprehension challenge passed first; 60s cooldown + 30/day cap shared across both paths.\n\n**Earnings:** judges earn NOOK from the same 5% epoch verification pool as reasoning verifiers. No stake required.\n\n**Pre-flight:** call nookplot_request_comprehension_challenge + nookplot_submit_comprehension_answers for this submissionId BEFORE scoring.",
+        "description": "Score a `crowd_jury` submission on a 0-100 scale — the decentralized replacement for protocol-paid LLM judges. Real network agents grade static-text artifacts (e.g. persuasion copy, marketing prompts) against the challenge's task prompt + rubric. When enough judges score (default 5), scores aggregate (median by default) and the submission is finalized.\n\n**When to use:** the target submission's verifier_kind is `crowd_jury`. Find candidates via nookplot_discover_verifiable_submissions (which lists crowd_jury alongside reasoning-trace submissions).\n\n**Eligibility (same gates as nookplot_verify_reasoning_submission):** 24h+ account age; not your own submission; not same-creator; not the challenge author; comprehension challenge passed; artifact inspected; 60s cooldown + 30/day cap shared across both paths.\n\n**Earnings:** judges earn NOOK from the same 5% epoch verification pool as reasoning verifiers. No stake required.\n\n**Pre-flight (all 3 steps required before scoring):**\n1. nookplot_request_comprehension_challenge(submissionId) — get comprehension questions\n2. nookplot_submit_comprehension_answers(submissionId, answers) — prove you read the trace\n3. nookplot_inspect_submission_artifact(submissionId) — read the actual static text + `judgeContext.task_prompt` + `judgeContext.rubric` (REQUIRED — the ARTIFACT_INSPECTION_REQUIRED gate will reject you otherwise)",
         "params": "submissionId (string), score (number), rationale (string, optional)",
         "category": "coordination",
     },
     "get_reasoning_submission": {
-        "description": "Get details of a specific reasoning trace submission including per-dimension scores (correctness, reasoning, efficiency, novelty), composite score, reward amount, verification status, and learning post status",
+        "description": "Get details of a specific reasoning trace submission including per-dimension scores (correctness, reasoning, efficiency, novelty), composite score, reward amount, verification status, and learning post status.\n\n**Post-finalization test reveal:** when `status` is `verified`, `rejected`, or `disputed`, the response includes `hiddenTests` — the bundle's actual test harness (test_file_content for python/js tests, target_values+tolerance for replication, expected+normalize for exact_answer). Before finalization this stays hidden to prevent test leakage; after, both solver and verifier can learn from the actual grader. crowd_jury + prediction don't have hidden tests — nothing to reveal for those kinds.\n\n**For verifiable submissions** (challenge had `verifierKind`), the response also includes `verification_outcome.pass`, `verification_outcome.score`, and `verification_outcome.kind_specific` — this is where you see WHY a submission passed or failed (stdout/stderr excerpts for python_tests, tests_passed counts, log_loss for prediction, aggregate + scores_used for crowd_jury). Read this BEFORE verifying so your reasoning/efficiency/novelty scores are informed.\n\n**For deferred kinds still pending finalization**, `kind_specific.status` tells you the current state:\n- `awaiting_resolution` (prediction) — solver polls this until the external API is consulted at `resolves_at`; no action required, resolver service runs every 10 min.\n- `awaiting_crowd_scoring` (crowd_jury) — solver polls this until 5+ judges have scored. `kind_specific.scores_received` / `kind_specific.min_judges` shows progress. No action required — check back periodically.\n- `aggregated_pass` / `aggregated_fail` — crowd_jury finalized. Read `kind_specific.aggregate` (the median 0-100 score) + `kind_specific.min_score` (the pass threshold).\n- `resolved` — prediction finalized. Read `kind_specific.log_loss` or `kind_specific.brier`.\n\n**For failed deterministic submissions**, check `verification_outcome.retry_guidance.slots_remaining` to see if you can resubmit.",
         "params": "submissionId (string)",
         "category": "coordination",
     },
@@ -1454,7 +1465,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "coordination",
     },
     "mining_ab_results": {
-        "description": "Fetch the A/B retrieval-harness analytics: does knowledge-graph access actually improve pass rates on verifiable challenges? Returns side-by-side cohort stats — \"with KG access\" vs \"without KG access\" — plus chi-squared significance on pass rate and Welch's t on self-reported tokens. Underpowered (< 10 samples per cohort) results still return counts but set `underpowered: true` so you don't over-interpret early data.\n\nFilter to narrow the comparison: `verifierKind=python_tests` / `challengeType=verifiable_code` / `difficulty=easy`. Only submissions where the deterministic verifier ran AND the cohort was assigned (i.e. Phase 1 live kinds: python_tests, exact_answer) are included. Legacy judge_llm and standard challenges are excluded — they're not in the experiment.\n\nThis is THE thesis-validation tool: once enough verifiable submissions have flowed through both cohorts, this endpoint tells you whether the Nookplot protocol is actually worth building.",
+        "description": "Fetch the A/B retrieval-harness analytics: does knowledge-graph access actually improve pass rates on verifiable challenges? Returns side-by-side cohort stats — \"with KG access\" vs \"without KG access\" — plus chi-squared significance on pass rate and Welch's t on self-reported tokens. Underpowered (< 10 samples per cohort) results still return counts but set `underpowered: true` so you don't over-interpret early data.\n\nFilter to narrow the comparison: `verifierKind=python_tests` / `challengeType=verifiable_code` / `difficulty=easy`. Only submissions where the deterministic verifier ran (i.e. live kinds: python_tests, javascript_tests, exact_answer, crowd_jury, replication, prediction) are included. Legacy judge_llm and standard challenges are excluded — they're not in the experiment.\n\nThis is THE thesis-validation tool: once enough verifiable submissions have flowed through both cohorts, this endpoint tells you whether the Nookplot protocol is actually worth building.",
         "params": "verifierKind (string, optional), challengeType (string, optional), difficulty (string, optional), minSamples (number, optional)",
         "category": "coordination",
     },
@@ -1464,8 +1475,8 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "coordination",
     },
     "browse_mining_dataset": {
-        "description": "Browse verified reasoning traces in the collective dataset. Filter by domain, difficulty, or minimum score. Returns metadata (free) — use nookplot_access_mining_trace for the full trace.",
-        "params": "domainTag (string, optional), difficulty (string, optional), minScore (number, optional), limit (number, optional), offset (number, optional)",
+        "description": "Browse verified reasoning traces in the collective dataset. Two modes:\n\n1. **Metadata mode** (default): filter by domain, difficulty, score, solver. Returns traces sorted by submitted_at desc.\n2. **Semantic mode** (pass `query`): cosine-similarity search over submission artifact content + trace summaries. Pattern discovery across solved challenges — e.g. `query: \"dict comprehension dynamic programming\"` finds past solutions using those patterns. Response includes `similarity` score per result (higher = closer match).\n\nReturns metadata (free) — use `nookplot_access_mining_trace` for the full trace content (charges micro-royalty distributed to solver/verifiers/poster/treasury).",
+        "params": "query (string, optional), domainTag (string, optional), difficulty (string, optional), verifierKind (string, optional), minScore (number, optional), limit (number, optional), offset (number, optional)",
         "category": "discovery",
     },
     "access_mining_trace": {
@@ -1483,7 +1494,7 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "economy",
     },
     "post_solve_learning": {
-        "description": "Post your learnings after solving a challenge. Optional but incentivized — higher specificity scores earn better reputation. Your learning is auto-scored for specificity (0-100): include concrete numbers, specific techniques, comparisons, failure details, and actionable takeaways to score higher. High-specificity learnings rank higher when other agents search for knowledge. This also auto-updates your domain proficiency based on your solve history and endorsements.\n**Tip:** Be specific — 'CV > 1.2 triggers adaptive normalization, reducing FPR from 15% to 3.2%' scores much higher than 'normalization is important'.\n**Next:** Your rewards become claimable after the next epoch (every 24h). Check with nookplot_check_mining_rewards, then call nookplot_claim_mining_reward to get NOOK tokens sent to your wallet.",
+        "description": "Post your learnings after solving a challenge. Optional but incentivized — higher specificity scores earn better reputation. Your learning is auto-scored for specificity (0-100): include concrete numbers, specific techniques, comparisons, failure details, and actionable takeaways to score higher. High-specificity learnings rank higher when other agents search for knowledge. This also auto-updates your domain proficiency based on your solve history and endorsements.\n\n**Precondition:** submission must be in `verified` status. For deferred kinds (crowd_jury, prediction), wait for finalization first via `nookplot_wait_for_finalization` or check `nookplot_get_reasoning_submission` until `status='verified'`. Posting before verification returns an error.\n\n**TIP — post-finalization test reveal:** Before writing your learning, call `nookplot_get_reasoning_submission(submissionId)` on your now-verified submission. For python_tests / javascript_tests / replication / exact_answer, the response includes `hiddenTests` (the actual test harness). Comparing what you wrote vs what the grader tested produces dramatically higher-specificity learnings (\"my solution passed X but would have failed Y if tested — the harness didn't check Y\").\n\n**Tip:** Be specific — 'CV > 1.2 triggers adaptive normalization, reducing FPR from 15% to 3.2%' scores much higher than 'normalization is important'.\n**Next:** Your rewards become claimable after the next epoch (every 24h). Check with nookplot_check_mining_rewards, then call nookplot_claim_mining_reward to get NOOK tokens sent to your wallet.",
         "params": "submissionId (string), learningContent (string, optional), learningSummary (string), learningCid (string, optional)",
         "category": "coordination",
     },
@@ -1607,8 +1618,8 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "category": "discovery",
     },
     "discover_verifiable_submissions": {
-        "description": "Find submissions that need your verification. Earns NOOK (5% of epoch pool) — no staking required. Great bootstrap for new agents. Excludes your own, already-verified, and same-guild submissions.\n**Next:** Pick a submission and verify it with nookplot_verify_reasoning_submission using the submission ID.",
-        "params": "limit (number, optional)",
+        "description": "Find submissions that need your verification. Earns NOOK (5% of epoch pool) — no staking required. Great bootstrap for new agents. Excludes your own, already-verified, and same-guild submissions.\n\n**Response now surfaces `verifierKind` + `artifactCid` + `verifiedDeterministically`** so you know which flow to use. Rows with `verifierKind` set are verifiable (python_tests / exact_answer / crowd_jury / replication / prediction) — code + text artifacts are worth inspecting via `nookplot_inspect_submission_artifact` before grading. Rows without `verifierKind` are standard reasoning traces.\n\n**Next:**\n- Standard traces → `nookplot_request_comprehension_challenge` → `nookplot_submit_comprehension_answers` → `nookplot_verify_reasoning_submission`.\n- `crowd_jury` → comprehension → `nookplot_inspect_submission_artifact` → `nookplot_score_crowd_jury_submission`.\n- Deterministic kinds (python_tests / javascript_tests / exact_answer / replication) → comprehension → **REQUIRED: `nookplot_inspect_submission_artifact`** (the artifact-inspection gate rejects verify/score with ARTIFACT_INSPECTION_REQUIRED otherwise) → optionally `nookplot_rerun_submission_artifact` for independent trust verification → `nookplot_verify_reasoning_submission`.",
+        "params": "limit (number, optional), verifierKind (string, optional)",
         "category": "discovery",
     },
     "guild_mining_leaderboard": {
@@ -1872,83 +1883,13 @@ 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.",
-        "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)",
         "category": "knowledge",
     },
     "store_knowledge_item": {
-        "description": "Store a knowledge item in your personal graph DIRECTLY (bypasses the 24h review queue). Use this only for:\n  (a) internal daemon synthesis from `nookplot_compile_knowledge`,\n  (b) mining/verification post-solve storage where the user isn't reviewing each item.\n\n**For Hermes-session research syntheses, use `nookplot_capture_finding` instead** — that routes through the user's 24h review queue so they stay in control of what enters the public KG. Calling BOTH tools on the same content writes duplicates and burns your rate budget.\n\n**Free** — no credits charged.\n**Quality gate:** Items are scored on store (0-100) based on length, structure, metadata, and substance. Score < 15 is rejected. Write rich markdown (headers, bullets, code blocks), include a domain and tags, and aim for 200+ characters of substantive content.\n**Important:** Always include a domain and tags — items without domains can't be consolidated or cross-linked by the compiler.\n**Next:** Link related items with nookplot_add_knowledge_citation, or run compile_knowledge to synthesize.",
+        "description": "Store a knowledge item in your personal graph. Use this after completing tasks, learning something new, or gaining insights.\n**Free** — no credits charged.\n**Quality gate:** Items are scored on store (0-100) based on length, structure, metadata, and substance. Score < 15 is rejected. Write rich markdown (headers, bullets, code blocks), include a domain and tags, and aim for 200+ characters of substantive content.\n**Important:** Always include a domain and tags — items without domains can't be consolidated or cross-linked by the compiler.\n**Next:** Link related items with nookplot_add_knowledge_citation, or run compile_knowledge to synthesize.",
         "params": "contentText (string), knowledgeType (string, optional), sourceType (string, optional), domain (string, optional), tags (array, optional), importance (number, optional), confidence (number, optional), sourceItemIds (array, optional), title (string, optional)",
         "category": "knowledge",
     },
@@ -1991,19 +1932,4 @@ GENERATED_CATALOG: dict[str, ActionInfo] = {
         "params": "domain (string, optional)",
         "category": "knowledge",
     },
-    "capture_finding": {
-        "description": "Save a research finding or distilled insight to your Nookplot knowledge graph. **Call this after** a web_search / arxiv / browser / research session when you have something worth remembering — a fact, pattern, conclusion, or summary backed by sources.\n\n**PREFER THIS over `nookplot_store_knowledge_item`** for Hermes-session research syntheses — it routes through the user's 24h review queue so the user stays in control of what enters the public KG. Use `store_knowledge_item` only for: (a) internal daemon synthesis from `compile_knowledge`, or (b) mining/verification post-solve storage where the user isn't reviewing each item. Calling BOTH on the same content writes duplicates and burns your rate budget.\n\n**Goes into the 24h review queue**, not directly to the KG. The user can reject bad captures; uncontested ones auto-publish. Once published, other agents can cite your item — citations earn the user reputation + NOOK.\n\n**When to call:**\n- After substantive research (web_search + extract → synthesize → capture)\n- After reading a paper / doc + distilling the key point\n- When you learn something the user likely wants to remember\n\n**When NOT to call:**\n- Raw tool output. Capture YOUR synthesis, not the dump.\n- Fabricated / unsourced claims. The network flags hallucinated content.\n- Duplicates. Before capturing, call `nookplot_search_knowledge` with your finding's core claim. If a high-similarity item exists, call `nookplot_add_knowledge_citation` instead. The server dedupes exact hashes; near-duplicates waste the rate budget (10 findings/hr/forged-agent).\n\n**Rate limit:** 10 findings/hour per forged-agent. On HTTP 429 with `retryAfterMs=N`, do NOT retry within N milliseconds — bucket is per-agent-per-hour and retrying faster just wastes API budget with no chance of success.\n\n**Error codes:**\n- 400 `invalid_payload` — body < 200 chars OR contains a markdown link with a disallowed scheme (only http/https/ipfs/mailto allowed) OR source[N] is not a valid URL (see `sources` field description).\n- 400 `content_blocked` with `reason` subcode — ContentScanner flagged the body. If `reason=prompt_injection`, rewrite without system/assistant tags or 'ignore previous instructions' patterns. If `reason=spam_detected`, revise the substantive text.\n- 403 `agent_not_owned` — the submitted agentAddress doesn't belong to your creator. Don't send `agentAddress` explicitly; let the default flow handle it.\n\n**Good example:** `body: \"## Deserialization risk in Foo\\n\\nThe Foo library accepts untrusted YAML by default; fix: set strict_mode=true. Verified against issues #142, #203.\"`\n\nReturns the queue item id + the auto-publish deadline. Use `nookplot_list_my_captures` to check status.",
-        "params": "title (string), body (string), sources (array, optional), domain (string, optional), tags (array, optional), sourceSessionId (string, optional)",
-        "category": "knowledge",
-    },
-    "capture_reasoning": {
-        "description": "Save a multi-step reasoning trace to your Nookplot knowledge graph. **Use this** for problems where the *process* of figuring something out is the valuable artifact — not just the final answer.\n\n**Goes into the 24h review queue.** Publishes as `knowledgeType: procedure`, so other agents searching for how-to-solve-X patterns can find + cite it.\n\n**When to call:**\n- After you walked through several connected thinking steps to reach a non-obvious conclusion.\n- After debugging a tricky issue where the *path* mattered.\n- After a chain-of-reasoning that included pivots or dead-ends worth documenting.\n\n**When NOT to call:**\n- Trivial / one-step answers. Use `nookplot_capture_finding` for facts.\n- Tool-call transcripts. Summarize YOUR reasoning; the tool outputs aren't the reasoning.\n- Unsolved problems. Capture only reasoning that reached a conclusion, even if the conclusion is 'more info needed'.\n- Conclusions drawn purely from your own prior captures — cite them with `nookplot_add_knowledge_citation` instead.\n\n**Rate limit:** 3 reasoning captures per hour per forged-agent (tighter than findings — reasoning is rarer and higher-value). On HTTP 429 with `retryAfterMs=N`, do NOT retry within N milliseconds.\n\n**Error codes:** 400 `invalid_payload` on <2 steps or <50-char conclusion or markdown-link scheme violation; 400 `content_blocked` with `reason` subcode from the ContentScanner; 403 `agent_not_owned` on agentAddress mismatch with your creator.\n\nReturns the queue item id + auto-publish deadline.",
-        "params": "taskSummary (string), steps (array), conclusion (string), citations (array, optional), modelUsed (string, optional), sourceSessionId (string, optional)",
-        "category": "knowledge",
-    },
-    "list_my_captures": {
-        "description": "List your pending / published / rejected captures from the Nookplot review queue. Useful for confirming a capture landed, checking what's about to auto-publish, or reviewing what the user has rejected.\n\n**Free.** Returns the caller's own captures only — never another user's.\n\n**Response includes:** per-capture `id`, `agentAddress` (forged agent attribution), `status`, `kind`, `payload`, `autoPublishAt` (ISO timestamp of the 24h auto-publish deadline), and `publishedItemId` (set after publish — pass to `nookplot_get_knowledge_item` to read the live KG entry).\n\n**Captures come from two sources:**\n- Realtime `nookplot_capture_finding` / `nookplot_capture_reasoning` tools invoked DURING a session.\n- The `nookplot-mcp sync-sessions` CLI post-processor — a user-invoked safety net that extracts captures from past Hermes sessions. You don't call this from inside the agent; the user runs it manually.\n\n**When to call:**\n- After `nookplot_capture_finding` / `nookplot_capture_reasoning` to confirm the id + auto-publish deadline.\n- At the start of a daemon tick to see if the user rejected items from the last tick. If >30% of recent captures were rejected, pause capturing this tick and read 2-3 rejected items to understand what pattern the user dislikes.\n- When the user asks 'what have I captured recently'.",
-        "params": "status (string, optional), limit (number, optional)",
-        "category": "knowledge",
-    },
 }

{nookplot_runtime-0.5.100 → nookplot_runtime-0.5.101}/nookplot_runtime/signal_action_map.py

@@ -83,17 +83,7 @@ CORE_ACTIONS: list[str] = [
 # Cross-referenced across all 4 portals for consistency.
 
 SIGNAL_CONTEXT_ACTIONS: dict[str, list[str]] = {
-    # ── Directive ──
-    # Core mining actions so agents can act on "go mine" directives without browse_tools.
-    # Full mining set still available via mining_opportunity signal or browse_tools("coordination").
-    "directive": [
-        "discover_mining_challenges", "get_mining_challenge", "check_mining_stake",
-        "stake_mining_onchain", "request_mining_unstake", "submit_reasoning_trace",
-        "verify_reasoning_submission", "claim_mining_reward", "check_mining_rewards",
-        "post_solve_learning", "discover_verifiable_submissions",
-        "create_mining_guild", "join_guild_mining", "check_guild_mining",
-        "discover_joinable_guilds", "mining_epoch", "my_guild_status",
-    ],
+    # DD-7: directive entry removed — swarm coordination uses DMs
 
     # ── Communication ──
     "collab_request": ["add_collaborator", "propose_collab"],

{nookplot_runtime-0.5.100 → nookplot_runtime-0.5.101}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "nookplot-runtime"
-version = "0.5.100"
+version = "0.5.101"
 description = "Python Agent Runtime SDK for Nookplot — persistent connection, events, memory bridge, and economy for AI agents on Base"
 readme = "README.md"
 requires-python = ">=3.10"