agentleFS
krzysukrzysu/market-skills/skills/l3-conviction-scan/SKILL.md

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.

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
  1. Very specific:

    After run-all-l3 to re-shape its per-ticker output into a single ranking.

    line 19
  2. Distinctive:

    Read them, don't strip them.

    line 77
  3. Very specific:

    L3 only: Does not run L2 patterns or load notes — those are run-watchlist's job.

    line 129

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

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_majors and crypto_alts first."
  • After run-all-l3 to 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-watchlist print.

When NOT to use

  • For per-ticker L2 + L3 + notes detail — use run-all-l3 (L3) or run-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 --narrative

Flags

FlagDefaultNotes
BASKET... (positional, repeatable)—One or more watchlist basket names.
--interval1d1m/5m/15m/30m/1h/2h/4h/8h/12h/1d/3d/1wk/1M.
--period1y1d/5d/1mo/3mo/6mo/1y/2y/5y/10y/ytd/max.
--sourceautoData provider override (yfinance, kraken, hl, ccxt:binance).
--watchlistskill defaultPath to watchlist.json (env: MARKET_SKILLS_WATCHLIST_PATH).
--top NunsetCap output rows to the top N by conviction.
--narrativeoffAfter the table, print the top-5 strategy narratives.
--jsonoffEmit 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-move

RRT2 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

  1. Resolve each positional basket via analysis.watchlist.by_category (the same market-watchlist resolver the other batch runners use).
  2. For each (basket, ticker), fetch_ohlc once, then call skills.run-all-l3.lib.analyze in-process with the watchlist's asset_class so scaled-maturity strategies (perp_dex, low_float, ai_infra) score correctly. No subprocess, no host-specific paths.
  3. Walk the envelope (tickers[t].strategies[s].ideas[]) and yield one flat row per idea. Tolerates both the canonical rr_to_tp: list[float] (preferred) and the legacy rr_to_tp2 / rr scalar fallback.
  4. 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.analyze via analysis.skill_loader.load_skill — never shells out, never reads host-specific config.
  • Asset-class-scaled conviction (perp_dex floor 3) means conv < 3 is 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.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.