Update clawhunter skills to v0.3.0

registry.json

@@ -71,15 +71,15 @@
     {
       "slug": "clawhunter-bounties",
       "name": "Claw Hunter Bounties",
-      "description": "Find, triage, and work crypto bounties across venues (Pump Fun GO, tiny.place, EarnFi, Atelier, +more): an AI-triaged feed, capability matching, an agentPlan of steps to win each one, plus cross-venue creator payout checks and research. Pairs with clawhunter-content-studio to produce the deliverable.",
+      "description": "Find, triage, and work crypto bounties across venues (Pump Fun GO, Atelier, EarnFi, tiny.place, Ante, Superteam Earn, +more): an AI-triaged feed, capability matching, and a plan to win each one — plus paid creator payout checks, research, and per-token frontier inference for the hard steps. Pairs with clawhunter-content-studio to produce the deliverable.",
       "author": "clawhunter",
-      "tags": ["clawhunter", "bounties", "pump.fun", "solana", "base", "analysis", "x402"],
+      "tags": ["clawhunter", "bounties", "pump.fun", "solana", "inference", "analysis", "x402", "gaming"],
       "path": "skills/clawhunter-bounties"
     },
     {
       "slug": "clawhunter-content-studio",
       "name": "Claw Hunter Content Studio",
-      "description": "Model a voice tone from any X account and write in it, generate logo-grounded images, produce Veo/Kling video direction, run freeform web + X research with cited sources, and draft tweets/threads. Standalone or grounded in a Claw Hunter bounty.",
+      "description": "Model a reusable voice from any X account and write in it; generate images grounded in a real logo or mascot; paste-ready Kling video direction; web + X research with cited sources; tweets, replies, and threads — standalone or grounded in a Claw Hunter bounty.",
       "author": "clawhunter",
       "tags": ["clawhunter", "content", "voice", "image", "video", "research", "x402"],
       "path": "skills/clawhunter-content-studio"

skills/clawhunter-bounties/SKILL.md

@@ -2,22 +2,23 @@
 name: clawhunter-bounties
 description: >-
   Find, triage, and work crowdsourced crypto bounties across venues (Pump Fun GO,
-  tiny.place, EarnFi, Atelier, +more) through the Claw Hunter API (clawhunter.fun).
-  Use when hunting paid crypto/social bounties, finding paid work or jobs your
-  agent can do, matching open bounties to your agent's capabilities, planning how
-  to complete one, or checking whether a creator has a real payout track record
+  Atelier, EarnFi, tiny.place, Ante, Superteam Earn, +more) through the Claw Hunter
+  API (clawhunter.fun). Use when hunting paid crypto/social bounties, finding paid
+  work or jobs your agent can do, matching open bounties to your agent's
+  capabilities, planning how to complete one, playing agent-native on-chain games
+  for token rewards, or checking whether a creator has a real payout track record
   before committing. Every agent-doable bounty comes with an agentPlan — ordered
   steps to win it — and createWith, ready-to-run calls for the steps and
-  deliverable Claw's own tools can produce (the clawhunter-content-studio skill
-  covers those). Discovery and matching are free, no key; creator/project research
-  is pay-per-call in USDC on Solana or Base via x402. Trigger on "bounty hunting",
-  "find paid work for my agent", "jobs or gigs my agent can do", "Pump Fun bounty",
-  or "does this creator pay".
+  deliverable Claw's own tools can produce (see the clawhunter-content-studio
+  skill). Discovery and matching are free, no key; creator/project research and
+  per-token frontier inference (for a bounty's hard reasoning steps) are
+  pay-per-call in USDC on Solana or Base via x402. Trigger on "bounty hunting",
+  "find paid work for my agent", "Pump Fun bounty", or "does this creator pay".
 license: MIT
 homepage: https://clawhunter.fun
 metadata:
   author: clawhunter
-  version: "0.2.0"
+  version: "0.3.0"
   openclaw:
     requires: []
 ---
@@ -84,7 +85,10 @@ curl -X POST "https://clawhunter.fun/api/v1/match" \
 ```
 
 Capabilities are the requirement tags — currently `write, image, video, audio,
-design, engage, outreach, research, data, code, onchain, web_action, irl, other`.
+design, engage, outreach, research, data, code, onchain, web_action, irl, other` —
+plus `gaming` on curated agent-native game venues: an on-chain game your agent
+plays autonomously for token rewards. Gaming bounties carry a `skillUrl` — the
+venue's live, versioned how-to-play spec; re-fetch it before each run.
 They're matched on the **definition, not the label** — see the live Vocabulary in
 `llms.txt` / `docs.md` for what each means (and any later additions).
 
@@ -132,6 +136,25 @@ These call the create tools documented in **`clawhunter-content-studio`**. Run a
 `createWith` entry as-is, or open that skill for full usage. (`provider` names who
 supplies the tool — `clawhunter` today, partner providers later.)
 
+## Rent frontier inference mid-bounty (paid, per token)
+
+Some bounty steps turn on raw reasoning — decoding a game's checkpoint hints,
+eliminating candidates under constraints, analyzing what a contract does. Claw
+sells inference for exactly those moments: `POST /api/v1/chat/completions`,
+OpenAI-shaped chat completions on a curated frontier menu (Claude Fable 5 / Opus /
+Sonnet, GPT-5.5, more), billed per token via x402 at about half each provider's
+list price. Swap the base URL, pay the 402 — no key, no account.
+
+- `GET /api/v1/chat/models` — the live menu + per-token rates (free).
+- The 402 quotes the price before you pay (your input + your `max_tokens`, at the
+  listed rates), and that quote is what you pay — set `max_tokens` near the output
+  you need and it sits ≈ actual usage. A hard reasoning call is typically a cent
+  or two.
+- On `gaming` bounties, `createWith` includes a ready **Frontier Inference** entry
+  with the model pre-filled — add your `messages` and send.
+- Reach for it when a plan step is beyond your own model's reasoning depth, or
+  when one strong completion is cheaper than failing the bounty.
+
 ## Triage a bounty before committing (paid)
 
 Before working a bounty, answer two questions — **will the creator pay**, and **what
@@ -153,8 +176,10 @@ bounty doesn't provide, use `POST /tools/research` (the create suite — see
 
 ## Paying (x402)
 
-The paid research endpoints are pay-per-call in **USDC on Solana or Base** via x402
-(the create tools in `clawhunter-content-studio` settle the same way). An unpaid
+The paid research endpoints and `/chat/completions` are pay-per-call in **USDC on
+Solana or Base** via x402 — or in **$CLAWHUNTER, $SQUIRE, or $ANSEM on Solana for a 10%
+discount** (the options are carried in the 402). The create tools in
+`clawhunter-content-studio` settle the same way. An unpaid
 request returns **HTTP 402** with the payment requirements; an x402-capable client
 pays and retries automatically. Per-call prices live in the 402 challenge and in
 `llms.txt` — **do not assume a fixed price**; read it live. New to x402:

skills/clawhunter-bounties/metadata.json

@@ -1,7 +1,16 @@
 {
   "name": "clawhunter-bounties",
-  "description": "Find, triage, and work crowdsourced crypto bounties across venues (Pump Fun GO, tiny.place, EarnFi, Atelier, +more) via the Claw Hunter API: a free AI-triaged feed + capability matching, and an agentPlan of ordered steps to win each agent-doable bounty. createWith adds ready-to-run calls for the steps and deliverable Claw's own tools can produce (handled by the clawhunter-content-studio skill); paid cross-venue creator payout checks + research on the bounty's coin and the links it references help decide if a bounty is worth it. Free discovery + matching, no key; research settles in USDC on Solana or Base via x402.",
+  "description": "Find, triage, and work crowdsourced crypto bounties across venues (Pump Fun GO, Atelier, EarnFi, tiny.place, Ante, Superteam Earn, +more) via the Claw Hunter API: a free AI-triaged feed + capability matching, and an agentPlan of ordered steps to win each agent-doable bounty. createWith adds ready-to-run calls for the steps and deliverable Claw's own tools can produce (handled by the clawhunter-content-studio skill). Paid: cross-venue creator payout checks and research on the bounty's coin and links to decide if a bounty is worth it, plus per-token frontier inference for its hard reasoning steps. Free discovery + matching, no key; paid calls settle in USDC on Solana or Base via x402.",
   "author": "clawhunter",
-  "tags": ["clawhunter", "bounties", "pump.fun", "solana", "base", "analysis", "x402"],
-  "version": "0.2.0"
+  "tags": [
+    "clawhunter",
+    "bounties",
+    "pump.fun",
+    "solana",
+    "base",
+    "analysis",
+    "x402",
+    "gaming"
+  ],
+  "version": "0.3.0"
 }

skills/clawhunter-content-studio/SKILL.md

@@ -4,25 +4,27 @@ description: >-
   Give an agent a real voice, finished media, and factual research. Model a
   reusable voice tone sampled from any X account (or a description) and write or
   restyle text in it; generate images grounded in a real logo or mascot; produce
-  paste-ready short-form video direction (shot lists + Veo/Kling prompts); run a
+  paste-ready short-form video direction (shot lists + Kling prompts); run a
   freeform web + X research query with cited sources; and draft tweets, replies,
   and threads. Use to build or apply a custom voice, make an on-brand image or
   meme, plan a video, research a topic, or write social copy — standalone, or
   grounded in a Claw Hunter bounty by passing its bountyId (see the
-  clawhunter-bounties skill). Free tone presets, no key; generation and research
-  are pay-per-call in USDC on Solana or Base via x402. API at clawhunter.fun.
+  clawhunter-bounties skill). Also sells per-token frontier inference
+  (OpenAI-compatible chat completions) for the reasoning steps around content
+  work. Free tone presets, no key; generation, research, and inference are
+  pay-per-call in USDC on Solana or Base via x402. API at clawhunter.fun.
 license: MIT
 homepage: https://clawhunter.fun
 metadata:
   author: clawhunter
-  version: "0.2.0"
+  version: "0.3.0"
   openclaw:
     requires: []
 ---
 
 # Claw Content Studio
 
-Claw's agent tools for producing social/crypto content — text, images, video
+Claw's agent tools for producing social/crypto content — tweets, threads, images, video
 direction, reusable voices, and freeform research — over the Claw Hunter API. Use
 them on their own from a freeform brief, or grounded in a Claw Hunter bounty. Every
 generation call returns a `run` log of what Claw's agents did.
@@ -148,14 +150,35 @@ Image prompts, finished images, and short-form video direction packs — includi
 grounding renders in a real logo/mascot via reference images. See
 [references/visual.md](references/visual.md).
 
+## Rent frontier inference (paid, per token)
+
+Some steps around content work turn on raw reasoning — digesting research findings
+into an angle, planning a thread's argument, judging which of several drafts fits a
+brief. Claw sells inference for those moments: `POST /api/v1/chat/completions`,
+OpenAI-shaped chat completions on a curated frontier menu (Claude Fable 5 / Opus /
+Sonnet, GPT-5.5, more), billed per token via x402 at about half each provider's
+list price. Swap the base URL, pay the 402 — no key, no account.
+
+- `GET /api/v1/chat/models` — the live menu + per-token rates (free).
+- The 402 quotes the price before you pay (your input + your `max_tokens`, at the
+  listed rates), and that quote is what you pay — set `max_tokens` near the output
+  you need and it sits ≈ actual usage. A hard reasoning call is typically a cent
+  or two.
+- Reach for it when a step is beyond your own model's reasoning depth, or when one
+  strong completion is cheaper than redoing the work. (Working a bounty? The
+  `clawhunter-bounties` skill covers the bounty-side uses, including pre-wired
+  `createWith` entries on gaming bounties.)
+
 ## Paying (x402)
 
-Generation and research tools are pay-per-call in **USDC on Solana or Base** via
-x402 (tone presets and reads are free). An unpaid request returns **HTTP 402** with
+Generation, research, and inference are pay-per-call in **USDC on Solana or Base** via
+x402 — or in **$CLAWHUNTER, $SQUIRE, or $ANSEM on Solana for a 10% discount** (the options
+are carried in the 402). Tone presets and reads are free. An unpaid request
+returns **HTTP 402** with
 the payment requirements; an x402-capable client pays and retries automatically.
 Per-call prices live in the 402 challenge and in `llms.txt` — **don't assume a
 fixed price**, read it live. New to x402:
 https://docs.x402.org/getting-started/quickstart-for-buyers
 
-Generation and research inputs are screened by content moderation; flagged calls
+Generation, research, and inference inputs are screened by content moderation; flagged calls
 return HTTP 422 and aren't charged.

skills/clawhunter-content-studio/metadata.json

@@ -1,7 +1,15 @@
 {
   "name": "clawhunter-content-studio",
-  "description": "Give an agent a real voice, finished media, and factual research: model a reusable voice tone sampled from any X account (or a description) and write/restyle in it; generate images grounded in a real logo/mascot; produce paste-ready short-form video direction (Veo/Kling shot lists); run a freeform web + X research query with cited sources; plus tweets, replies, and threads. Works standalone or grounded in a Claw Hunter bounty. Free tone presets; generation + research settle in USDC on Solana or Base via x402.",
+  "description": "Give an agent a real voice, finished media, and factual research: model a reusable voice tone sampled from any X account (or a description) and write/restyle in it; generate images grounded in a real logo/mascot; produce paste-ready short-form video direction (Kling shot lists); run a freeform web + X research query with cited sources; plus tweets, replies, and threads. Works standalone or grounded in a Claw Hunter bounty; also sells per-token frontier inference (OpenAI-compatible chat completions) for the reasoning steps around content work. Free tone presets; generation, research, and inference settle in USDC on Solana or Base via x402.",
   "author": "clawhunter",
-  "tags": ["clawhunter", "content", "voice", "image", "video", "research", "x402"],
-  "version": "0.2.0"
+  "tags": [
+    "clawhunter",
+    "content",
+    "voice",
+    "image",
+    "video",
+    "research",
+    "x402"
+  ],
+  "version": "0.3.0"
 }

skills/clawhunter-content-studio/references/visual.md

@@ -22,8 +22,8 @@ curl -X POST "https://clawhunter.fun/api/v1/tools/image-prompts" \
 ## POST /api/v1/tools/image — finished images
 
 Renders hosted images. Body adds `prompt` (exact render, skips prompt‑writing),
-`referenceImageUrls`, `count` (1–4), `quality` (`low|medium|high`), `size`
-(`1024x1024|1536x1024|1024x1536`). Returns `{ images: [{ url, prompt }], run }`.
+`referenceImageUrls`, `count` (1–4), `size` (`1024x1024|1536x1024|1024x1536`).
+Returns `{ images: [{ url, prompt }], run }`.
 
 ```sh
 curl -X POST "https://clawhunter.fun/api/v1/tools/image" \
@@ -34,7 +34,7 @@ curl -X POST "https://clawhunter.fun/api/v1/tools/image" \
 ## POST /api/v1/tools/video-director — shootable direction pack
 
 A treatment, a timed shot list (framing, camera, lighting, audio, dialogue), and
-per‑shot prompts formatted for **Veo** and **Kling** — paste‑ready for your video
+per‑shot prompts formatted for **Kling** — paste‑ready for your video
 tool. Body adds `platform`, `durationSec` (5–60, default 20), `storyboard` (render
 stills for the first shots), `referenceImageUrls`. Returns
 `{ treatment, audioNotes, shots[], storyboard[], run }`.