Add solana-bounty-operator-skill

solana-bounty-operator-skill/README.md

@@ -0,0 +1,58 @@
+# solana-bounty-operator-skill
+
+An agent skill for safely finding, triaging, and preparing submissions for Solana ecosystem bounties.
+
+The skill is designed for AI coding agents that operate around Superteam, GitHub PR bounties,
+hackathons, and grant-style opportunities. Its job is not to spam submissions. Its job is to decide
+whether a task is eligible, whether the agent can honestly complete it, what evidence is required,
+and which submission channel is allowed.
+
+## Why This Skill Exists
+
+Solana builders now discover work across Superteam Earn, GitHub issues, hackathons, and partner
+grant pages. Agents can help, but without a safety workflow they easily make costly mistakes:
+
+- submitting after deadline
+- ignoring region or KYC restrictions
+- duplicating a previous submission
+- claiming social reach or wallet ownership they cannot prove
+- publishing low-evidence work that cannot be reviewed
+- calling paid APIs or signing wallets without explicit approval
+
+This skill gives agents a repeatable operating procedure for bounty work.
+
+## What It Does
+
+- Reads a listing or issue and extracts deadline, region, reward, required links, account needs, and
+  accepted submission channels.
+- Classifies the opportunity as `agent-submit`, `human-submit`, `build-first`, `skip`, or `ask-user`.
+- Produces a deliverable plan with evidence requirements.
+- Builds a submission packet with links, answers, and reviewer notes.
+- Blocks unsafe actions such as fake region claims, fabricated followers, wallet signatures, and
+  duplicated submissions.
+
+## Files
+
+- `skill/SKILL.md` - main agent instructions
+- `skill/sources.md` - how to inspect Superteam/GitHub opportunities
+- `skill/submission-safety.md` - hard stop rules
+- `scripts/triage-listing.mjs` - deterministic listing triage helper
+- `examples/superteam-listing.example.json` - sample input
+- `examples/triage-output.example.json` - sample output
+- `commands/triage-bounty.md` - slash-command style workflow
+- `agents/bounty-operator.md` - suggested agent persona
+- `validate.sh` - local smoke validation
+
+## Quick Check
+
+```bash
+cd solana-bounty-operator-skill
+./validate.sh
+node scripts/triage-listing.mjs examples/superteam-listing.example.json
+```
+
+## Fit With Solana AI Kit
+
+This skill complements coding and Solana transaction skills by handling the work acquisition layer:
+which bounty to work on, whether the agent may submit, and what proof must be attached. It is useful
+for autonomous Solana contributors, hackathon agents, and teams running bounty pipelines.

solana-bounty-operator-skill/agents/bounty-operator.md

@@ -0,0 +1,24 @@
+# Bounty Operator Agent
+
+Persona for an autonomous bounty operator.
+
+## Mission
+
+Find real Solana ecosystem opportunities, complete only the ones that can be proven, and submit
+through the correct channel without misrepresenting the human.
+
+## Behavior
+
+- Skeptical about eligibility.
+- Fast at building small artifacts.
+- Strict about deadlines.
+- Keeps a local ledger of submissions.
+- Does not optimize for spam volume.
+
+## Success Metrics
+
+- Valid submissions accepted by the platform.
+- No secret exposure.
+- No fake social or region claims.
+- Clear evidence links.
+- Fewer duplicate or low-quality submissions.

solana-bounty-operator-skill/commands/triage-bounty.md

@@ -0,0 +1,33 @@
+# /triage-bounty
+
+Use this command when the user asks to find, evaluate, or submit Solana bounties.
+
+## Input
+
+- Listing or issue URL
+- Optional user-provided accounts and eligibility facts
+
+## Steps
+
+1. Fetch the source.
+2. Extract facts into JSON.
+3. Run `scripts/triage-listing.mjs` if the source is a structured listing.
+4. Apply `skill/submission-safety.md`.
+5. Return a decision:
+   - `agent-submit`
+   - `human-submit`
+   - `build-first`
+   - `skip`
+   - `ask-user`
+6. If `agent-submit`, create the deliverable and submit.
+7. Record submission ID or PR URL.
+
+## Response Template
+
+```text
+Decision:
+Why:
+Work I can do now:
+User action needed:
+Submission/evidence link:
+```

solana-bounty-operator-skill/examples/superteam-listing.example.json

@@ -0,0 +1,10 @@
+{
+  "title": "Create Twitter Posts Explaining Streamflow Business",
+  "slug": "create-twitter-posts-explaining-streamflow-business",
+  "sponsor": { "name": "Streamflow Finance" },
+  "deadline": "2026-07-17T21:59:59.999Z",
+  "region": "Global",
+  "agentAccess": "AGENT_ALLOWED",
+  "isWinnersAnnounced": false,
+  "requiredDeliverables": ["X thread link"]
+}

solana-bounty-operator-skill/examples/triage-output.example.json

@@ -0,0 +1,10 @@
+{
+  "title": "Create Twitter Posts Explaining Streamflow Business",
+  "slug": "create-twitter-posts-explaining-streamflow-business",
+  "sponsor": "Streamflow Finance",
+  "deadline": "2026-07-17T21:59:59.999Z",
+  "region": "Global",
+  "agentAccess": "AGENT_ALLOWED",
+  "decision": "agent-submit",
+  "reason": "agent submission channel allowed"
+}

solana-bounty-operator-skill/scripts/triage-listing.mjs

@@ -0,0 +1,51 @@
+#!/usr/bin/env node
+import { readFileSync } from 'node:fs'
+
+const file = process.argv[2]
+if (!file) {
+  console.error('usage: node scripts/triage-listing.mjs <listing.json>')
+  process.exit(2)
+}
+
+const listing = JSON.parse(readFileSync(file, 'utf8'))
+const now = new Date(process.env.NOW_ISO || new Date().toISOString())
+
+function deadlineState(value) {
+  if (!value) return { expired: false, reason: 'no deadline supplied' }
+  const d = new Date(value)
+  if (Number.isNaN(d.getTime())) return { expired: true, reason: 'invalid deadline' }
+  return d < now
+    ? { expired: true, reason: `deadline passed at ${d.toISOString()}` }
+    : { expired: false, reason: `deadline open until ${d.toISOString()}` }
+}
+
+function decide(x) {
+  const deadline = deadlineState(x.deadline)
+  if (deadline.expired) return { decision: 'skip', reason: deadline.reason }
+  if (x.isWinnersAnnounced) return { decision: 'skip', reason: 'winners already announced' }
+  if (x.region && !['Global', ''].includes(x.region) && !x.userRegionConfirmed) {
+    return { decision: 'ask-user', reason: `region is ${x.region}; user eligibility not confirmed` }
+  }
+  if (x.agentAccess === 'HUMAN_ONLY') {
+    return { decision: 'human-submit', reason: 'listing requires human account submission' }
+  }
+  if (Array.isArray(x.requiredDeliverables) && x.requiredDeliverables.some((d) => /video|demo|repo|deck/i.test(d))) {
+    return { decision: 'build-first', reason: 'requires a real artifact before submission' }
+  }
+  if (x.agentAccess === 'AGENT_ALLOWED' || x.agentAccess === 'AGENT_ONLY') {
+    return { decision: 'agent-submit', reason: 'agent submission channel allowed' }
+  }
+  return { decision: 'ask-user', reason: 'submission channel unclear' }
+}
+
+const result = {
+  title: listing.title,
+  slug: listing.slug,
+  sponsor: listing.sponsor?.name || listing.sponsor,
+  deadline: listing.deadline,
+  region: listing.region,
+  agentAccess: listing.agentAccess,
+  ...decide(listing),
+}
+
+console.log(JSON.stringify(result, null, 2))

solana-bounty-operator-skill/scripts/validate.mjs

@@ -0,0 +1,33 @@
+#!/usr/bin/env node
+import { existsSync, readFileSync } from 'node:fs'
+import { execFileSync } from 'node:child_process'
+
+const required = [
+  'README.md',
+  'skill/SKILL.md',
+  'skill/sources.md',
+  'skill/submission-safety.md',
+  'scripts/triage-listing.mjs',
+  'examples/superteam-listing.example.json',
+]
+
+for (const file of required) {
+  if (!existsSync(file)) throw new Error(`missing required file: ${file}`)
+}
+
+const out = execFileSync(
+  process.execPath,
+  ['scripts/triage-listing.mjs', 'examples/superteam-listing.example.json'],
+  { encoding: 'utf8' },
+)
+const triage = JSON.parse(out)
+if (triage.decision !== 'agent-submit') {
+  throw new Error(`expected agent-submit, got ${triage.decision}`)
+}
+
+const skill = readFileSync('skill/SKILL.md', 'utf8')
+for (const phrase of ['Hard Stops', 'Superteam Notes', 'GitHub Bounty Notes']) {
+  if (!skill.includes(phrase)) throw new Error(`SKILL.md missing section: ${phrase}`)
+}
+
+console.log('solana-bounty-operator-skill validation passed')

solana-bounty-operator-skill/skill/SKILL.md

@@ -0,0 +1,96 @@
+---
+name: solana-bounty-operator
+description: Safely discover, triage, prepare, and submit Solana ecosystem bounty/grant work without fabricating eligibility or using restricted user actions.
+---
+
+# Solana Bounty Operator
+
+Use this skill when an agent is asked to find or submit Solana ecosystem bounties, grants, hackathon
+tracks, GitHub issue bounties, or Superteam Earn opportunities.
+
+## Primary Rule
+
+Submit only when the agent has a truthful, reviewable deliverable and the submission channel allows
+agent or delegated submission. If the opportunity requires a human social account, KYC, wallet
+signature, paid call, private credential, or region claim, stop and ask for explicit human action.
+
+## Workflow
+
+1. Capture the source URL and current timestamp.
+2. Extract opportunity facts:
+   - title
+   - sponsor
+   - reward and token
+   - deadline
+   - region
+   - required deliverables
+   - submission channel
+   - account requirements
+   - wallet or payment requirements
+   - judging criteria
+3. Classify the opportunity:
+   - `agent-submit`: agent API, GitHub PR, or public form accepts agent/delegated submission.
+   - `human-submit`: requires X, Discord, Telegram, KYC, region proof, or a logged-in user account.
+   - `build-first`: requires code, video, deck, demo, or research before submission.
+   - `skip`: expired, region-ineligible, duplicate, or impossible to verify.
+   - `ask-user`: missing required identity, wallet, or account consent.
+4. Build only real deliverables.
+5. Create a submission packet:
+   - main link
+   - evidence links
+   - eligibility answers
+   - reviewer note
+   - safety note listing anything not performed
+6. Submit through the allowed channel.
+7. Record:
+   - submission ID or PR URL
+   - status
+   - exact links sent
+   - deadline / winners date
+   - actions not performed
+
+## Hard Stops
+
+Read `submission-safety.md` before submitting.
+
+Never:
+
+- claim followers, geography, citizenship, KYC status, or community role not provided by the user
+- sign a wallet transaction without explicit approval
+- make paid API calls without explicit approval
+- submit a duplicate packet to the same listing unless updating the existing submission is allowed
+- invent demo links or publish empty repos
+- submit after the deadline unless the sponsor explicitly says late submissions are accepted
+
+## Superteam Notes
+
+If using Superteam:
+
+- Prefer an authenticated agent endpoint only when `agentAccess` permits it.
+- If the public listing says `HUMAN_ONLY`, do not bypass it with an agent.
+- For projects, include Telegram only if the human provided it.
+- For bounties, the `link` field should be the primary deliverable URL.
+- `eligibilityAnswers` must match each required question exactly.
+
+## GitHub Bounty Notes
+
+If the allowed submission channel is GitHub:
+
+- Fork the repo.
+- Create a narrowly scoped branch.
+- Add reproducible files and validation.
+- Open a PR with proof, run commands, and no secrets.
+- Link any standalone demo or docs.
+
+## Output Format
+
+When reporting back:
+
+```text
+Opportunity:
+Decision:
+Reason:
+Deliverable:
+Submission:
+Remaining user action:
+```

solana-bounty-operator-skill/skill/sources.md

@@ -0,0 +1,62 @@
+# Sources And Extraction
+
+## Superteam Earn
+
+Useful fields:
+
+- `title`
+- `slug`
+- `type`
+- `rewardAmount`
+- `token`
+- `deadline`
+- `region`
+- `agentAccess`
+- `eligibility`
+- `applicationType`
+- `compensationType`
+- `pocSocials`
+- `isWinnersAnnounced`
+
+Decision hints:
+
+- `AGENT_ALLOWED` or `AGENT_ONLY`: agent submission may be possible.
+- `HUMAN_ONLY`: prepare material only; human must submit.
+- `region` other than `Global`: verify the human is eligible before submitting.
+- `eligibility` with social reach questions: never invent metrics.
+- `type=project`: usually needs Telegram and sometimes compensation ask.
+
+## GitHub Issue / PR Bounties
+
+Extract:
+
+- repo owner/name
+- issue or bounty number
+- acceptance criteria
+- expected branch/PR target
+- tests
+- payout instructions
+- whether work is first-come or judged
+
+Decision hints:
+
+- A PR can be agent-submitted if the repo accepts public contributions.
+- A claim issue may require maintainers to assign the bounty first.
+- If payout requires a wallet, ask the human before posting it publicly.
+
+## Hackathons
+
+Extract:
+
+- exact track
+- deadline
+- required links
+- video/deck length
+- demo requirements
+- team/account requirements
+- judging criteria
+
+Decision hints:
+
+- If the demo is not real, classify as `build-first`.
+- If a live wallet transaction is required, ask for approval and use devnet where possible.