coin-narrative-catcher

CT-first memecoin narrative analysis skill for OpenClaw.

Status: ✅ CI passing on main.

What it does

• Prioritizes catalyst origin + credibility graph + narrative payload
• Uses tape/flow as secondary confirmation
• Includes launchpad attribution checks for:
  • Virtuals (exact)
  • Clanker (exact)
  • Bankr (exact)
  • Flaunch (heuristic)
  • Doppler (heuristic/index signal)

What it does not do

• place trades
• guarantee a token is safe or legitimate
• treat all launchpad signals as equally strong
• replace primary-source verification for identity / prestige claims

Architecture

CA / token input
    ↓
launchpad attribution
    ↓
creator / linked-handle discovery
    ↓
credibility graph
    ↓
CT narrative read
    ↓
confidence tier + final report

Trust model (attribution confidence)

• exact: first-party launch source returns an exact CA match
• heuristic: strong indirect evidence (e.g., flETH quote token, indexed launch page)
• none: no reliable source signal

Output includes:

• classification
• confidenceTier
• alsoIndexedOn (secondary index hints)
• linked handles (current), creator email pattern hints, and required links
• aGDP deep link when available (https://agdp.io/agent/{id}), otherwise explicit manual-lookup fallback
• directionality guardrail reminder (verify A→B and B→A separately)

Why these APIs are used

These are not random outbound calls; they are domain-specific crypto launch/trading data providers.

• api.dexscreener.com for market/tape and quote-token heuristics
• api2.virtuals.io for Virtuals attribution
• www.clanker.world for Clanker exact-match attribution
• api.bankr.bot for Bankr launch metadata (including deployer + fee recipient)
• app.doppler.lol for Doppler index detection
• api.fxtwitter.com for tweet resolution where needed
• agdp.io for Virtuals agent deep links

See docs/data-sources.md for detailed trust/failure notes.

Structure

• SKILL.md — core skill instructions
• references/ — scoring framework, output template, attribution rules
• scripts/check_launchpads.py — classify CA across launchpad sources
• scripts/credibility_graph.py — build lightweight identity/credibility graph
• scripts/analyze_narrative.py — one-command narrative draft report
• tests/ — attribution logic tests

Quick usage

Launchpad attribution

python3 scripts/check_launchpads.py <CA>

Credibility graph

python3 scripts/credibility_graph.py --handles <x_handle1> <x_handle2> --github <gh_user1>

One-command narrative draft

python3 scripts/analyze_narrative.py --ca <CA> --x-handle <key_handle>

Example report shape

classification: Virtuals
confidenceTier: exact
alsoIndexedOn: []
linkedHandles: [project_handle, creator_handle]
keyQuestion: who launched this, who is pushing it, and why does the market care?
read: narrative looks strong / mixed / weak depending on catalyst + credibility graph

The point is not to output a magic score in a vacuum. The point is to reduce a memecoin into the few identity and catalyst questions that actually matter.

Quality controls

• CI: .github/workflows/ci.yml
• Tests: python -m unittest discover -s tests -v
• Security policy: SECURITY.md
• Contribution guide: CONTRIBUTING.md
• License: LICENSE (MIT)

Maintainer

• X: @solgoodman

Disclaimer

This repository is a research and analysis helper. It is not financial advice, not an execution engine, and not a guarantee of token legitimacy. Always verify independently before trading.

Versioning / releases

Use semantic versioning tags:

• v0.1.0 initial stable milestone
• v0.1.1 bugfix-only release
• v0.2.0 new capability release (e.g., new launchpad support)

Recommended flow:

git tag v0.1.0
git push origin v0.1.0

Then create a GitHub Release with a short changelog.

Package skill

python3 ~/.npm-global/lib/node_modules/openclaw/skills/skill-creator/scripts/package_skill.py . ./dist