Add gemini-extension.json manifest with correct array-format settings,
symlink skills/last30days/SKILL.md to root SKILL.md for Gemini skill
discovery, add Gemini install paths to bash for-loop in both main and
open variant, and add Gemini CLI install instructions to README.
Incorporates the good parts of PR #53 (manifest, paths, README) while
avoiding duplicate SKILL.md, tool name scattering, and allowed-tools
pollution that would have created maintenance issues.
Closes#45
Co-Authored-By: Alex Ferrari <alex@thealexferrari.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- _compute_text_similarity() now checks outcome names with bidirectional
substring matching (0.85) and token overlap (0.7), not just event titles
- Collect outcomes from ALL active markets per event, filter to >1% price
- Reorder outcome_prices to surface topic-matching outcome before top-3 truncation
- Add SKILL.md "Prediction Markets" synthesis section with structural/long-term
market preference, domain examples, citation format, and narrative weaving
- Add Polymarket to citation priority list between HN and Web
- Update stats box template to show up to 5 market odds
- Fix render.py "vol24h" label to "volume"
- Add NCAA seed fixture event for outcome-only matching tests
- 82 polymarket tests pass (14 new)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Make the "omit 0-result sources" instruction more emphatic (CRITICAL prefix,
enumerate all formats to suppress). Change Polymarket stats line to show
top 2-3 market odds instead of just top-1.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Search Polymarket's free Gamma API for relevant prediction markets on any
topic. Uses smart multi-query expansion to cast a wider net (e.g., "Arizona
Basketball" also searches "Arizona"), merges and dedupes by event ID, and
shows price movement context ("up 22.5% this week"). No API key required.
Also hides sources with zero results from the stats output (all sources).
54 new tests, all passing. Full pipeline integration with scoring, dedupe,
cross-source linking, and rendering.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Handle resolution was only triggering for "person/brand" topics. Products
and tools like "Nano Banana Pro" or "Seedance" can also have X accounts
(@nanobanana, @seedanceai) but were being skipped.
Changes:
- Broadened trigger: people, brands, products, tools, companies, communities
- Added parody/fan account verification guidance
- Added site:x.com to WebSearch query for better results
- Updated both SKILL.md and OpenClaw variant
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When a topic is a person/brand (e.g. "Dor Brothers", "Jason Calacanis"),
the agent now resolves their X handle via WebSearch before running the
script, then passes --x-handle to search their posts unfiltered (no
topic keywords required). This finds posts the entity made without
mentioning their own name.
- SKILL.md + OpenClaw variant: Step 0.5 handle resolution instructions
- last30days.py: --x-handle CLI arg, passed through to _run_supplemental()
- bird_x.search_handles(): topic is now Optional[str] for unfiltered mode
- schema.py: resolved_x_handle field on Report
- render.py: show resolved handle in stats output
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace CLI-style promo banners with casual, conversational NUX message
that teaches users to say "last30" + natural language. Remove stale Bird
CLI npm install references and dead prompt_bird_install methods.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Matt Van Horn