l3-conviction-scan
Skill3 starsChanged 4 months agoMITView on GitHub
Conviction-ranked L3 cross-ticker view. Flattens every L3 idea across one or more watchlist baskets, sorts by conviction, and surfaces the top N. Use for morning briefs, conviction triage, or any LLM scan that needs 'show me the strongest setups first' across the whole watchlist.
In this repository47 other files it gives its agents.
- AGENTS.mdAGENTS.md
- beadsSkill
- backtest-engineSkill
- backtest-pipelineSkill
- backtest-trend-minerSkill
- bug-scanSkill
- daily-trade-pickSkill
- execution-kraken-perpsSkill
- execution-kraken-spotSkill
- market-accumulationSkill
- market-basisSkill
- market-breadthSkill
- market-breakoutSkill
- market-emaSkill
- market-exhaustionSkill
- market-fibonacciSkill
- market-liquidity-sweepSkill
- market-macdSkill
- market-macroSkill
- market-moversSkill
- market-notesSkill
- market-overviewSkill
- market-rsiSkill
- market-snapshotSkill
- market-squeezeSkill
- market-s-rSkill
- market-stateSkill
- market-trend-qualitySkill
- market-trendSkill
- market-valuationSkill
- market-volatilitySkill
- market-volumeSkill
- market-watchlistSkill
- portfolio-mgmtSkill
- position-watchdogSkill
- risk-engineSkill
- run-all-l2Skill
- run-all-l3Skill
- run-watchlistSkill
- strategy-accumulation-swingSkill
- strategy-breakout-confirmSkill
- strategy-exhaustion-fadeSkill
- strategy-fitness-heatmapSkill
- strategy-funding-carrySkill
- strategy-liquidity-sweepSkill
- strategy-mean-reversionSkill
- strategy-trend-followSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
After run-all-l3 to re-shape its per-ticker output into a single ranking.
- Distinctive: line 77
Read them, don't strip them.
- Very specific: line 129
L3 only: Does not run L2 patterns or load notes — those are run-watchlist's job.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
l3-conviction-scan
Ranks L3 trade ideas across watchlist baskets by conviction. Fetches candles once per ticker per (basket, interval, period), runs all 7 L3 strategies in-process, then flattens every ideas[] into a single conviction-sorted table.
When to use
- Morning briefs and swing scans: "show me the strongest setups across
crypto_majorsandcrypto_altsfirst." - After
run-all-l3to re-shape its per-ticker output into a single ranking. - Any LLM-driven triage that needs a flat "best N trades right now" view, not the
per-ticker breakdown that
run-all-l3/run-watchlistprint.
When NOT to use
- For per-ticker L2 + L3 + notes detail — use
run-all-l3(L3) orrun-watchlist(L2 + L3 + notes). This skill is a ranking view on top of L3, not a replacement. - For diagnostic / anomaly detection on the L2 envelope — use
bug-scan. - For execution — never pipe conviction-scan output to
execution-kraken-*without re-running the L3 strategy's full narrative + risk vet.
Quick Start
# All ideas across crypto_majors + crypto_alts, default 1d/1y
uv run skills/l3-conviction-scan/scripts/run.py crypto_majors crypto_alts
# Top 10, intraday TF, machine-readable
uv run skills/l3-conviction-scan/scripts/run.py crypto_alts --interval 4h --period 3mo --top 10 --json
# Include top-5 strategy narratives after the table
uv run skills/l3-conviction-scan/scripts/run.py crypto_majors crypto_alts --narrativeFlags
| Flag | Default | Notes |
BASKET... (positional, repeatable) | — | One or more watchlist basket names. |
--interval | 1d | 1m/5m/15m/30m/1h/2h/4h/8h/12h/1d/3d/1wk/1M. |
--period | 1y | 1d/5d/1mo/3mo/6mo/1y/2y/5y/10y/ytd/max. |
--source | auto | Data provider override (yfinance, kraken, hl, ccxt:binance). |
--watchlist | skill default | Path to watchlist.json (env: MARKET_SKILLS_WATCHLIST_PATH). |
--top N | unset | Cap output rows to the top N by conviction. |
--narrative | off | After the table, print the top-5 strategy narratives. |
--json | off | Emit machine-readable JSON envelope to stdout. |
Both --interval and --period are validated via analysis.intervals.validate_timeframe — bad values exit 2 with a friendly error.
Output
Text (default)
TF BASKET TICKER STRATEGY DIR CONV ENTRY STOP TP1 TP2 RRT2 VETO
---- --------------- ---------- -------------------------- ----- ---- --------- --------- --------- --------- ------ -----
1d crypto_majors BTCUSD strategy-trend-follow long 5 67500.00 66200.00 70500.00 73000.00 2.44
1d crypto_majors ETHUSD strategy-breakout-confirm long 4 187.20 182.50 194.00 199.50 2.31
4h crypto_alts HYPEUSD strategy-trend-follow long 4 24.10 22.80 26.40 28.10 2.29
1d crypto_alts SOLUSD strategy-mean-reversion long 3 3450.00 3380.00 3550.00 3620.00 1.78 mature-moveRRT2 is the deployment-ready R:R to TP2 (the runner TP1 here is just the
scaling-out target; see run-all-l3 output schema). The VETO column
shows soft veto tags emitted by L3 strategies (mature-move,
asset-class-scaled, etc.) — these are correct L3 behaviour, not bugs.
Read them, don't strip them.
JSON
{
"interval": "1d",
"period": "1y",
"baskets": ["crypto_majors", "crypto_alts"],
"count": 42,
"ideas": [
{
"ticker": "BTCUSD",
"label": "Bitcoin",
"tier": 2,
"asset_class": "majors",
"strategy": "strategy-trend-follow",
"direction": "long",
"conviction": 5,
"version": "v5",
"entry": 67500.0,
"stop": 66200.0,
"tp1": 70500.0,
"tp2": 73000.0,
"tp3": 76000.0,
"rr_tp1": 1.22,
"rr_tp2": 2.44,
"rr_tp3": 3.94,
"mover_pct": 18.4,
"veto": [],
"narrative": "Healthy uptrend with higher-low structure..."
}
]
}How it works
- Resolve each positional basket via
analysis.watchlist.by_category(the samemarket-watchlistresolver the other batch runners use). - For each (basket, ticker),
fetch_ohlconce, then callskills.run-all-l3.lib.analyzein-process with the watchlist'sasset_classso scaled-maturity strategies (perp_dex, low_float, ai_infra) score correctly. No subprocess, no host-specific paths. - Walk the envelope (
tickers[t].strategies[s].ideas[]) and yield one flat row per idea. Tolerates both the canonicalrr_to_tp: list[float](preferred) and the legacyrr_to_tp2/rrscalar fallback. - Sort by
conviction desc(ties broken by ticker), optionally cap to--top N, render.
Layer rules
- L3 only. Does not run L2 patterns or load notes — those are
run-watchlist's job. Conviction scan is the cheap, fast "what's the top N?" view. - Advisory only. Like every other batch runner, conviction-scan produces a view; it does not gate trades. The risk engine + execution prompt remain the safety layer.
- In-process. Imports
skills.run-all-l3.lib.analyzeviaanalysis.skill_loader.load_skill— never shells out, never reads host-specific config. - Asset-class-scaled conviction (perp_dex floor 3) means
conv < 3is not a defect, just a veto signal in the swing-scan playbook. This skill surfaces everything; downstream agents decide what to act on.
Output envelope (AXI)
--json output follows the canonical AXI envelope — {data, count, errors, help[]}. Pass --fields=<csv> to project or --full for the full payload. count is the item count (findings for bug-scan, ranked ideas for l3-conviction-scan, total journal entries for daily-trade-pick), help[] is contextual next-step command templates.
Home view (no-arg mode)
No-arg mode prints the home view from $XDG_DATA_HOME/market-skills/<skill>_last.json (last successful run). Errors are not cached.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

