agentleFS
INo-xiousINo-xious/stockbit-mcp/skills/watch/SKILL.md

watch

Skill58 starsChanged 3 months agoMITView on GitHub

Watch live IDX turnover and order books for unusual activity — value surges, auto-reject approach, one-sided depth, walls appearing or withdrawn, quiet stocks waking up, UMA flags — plus delayed print-by-print attribution and end-of-day broker context. Use when the user runs /watch, or asks what is trading right now, who is moving, whether anyone is dumping, or where the money is going in the Indonesian market.

48 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use decision-map, map and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Mathews-TomSkill

    decision-map

    Use when mapping unresolved architecture, policy, or scope decisions before planning, or working one decision ticket, including a prototype ticket whose throwaway evidence the prototype package builds. Also applies to "map the decisions", "decision map", and "wayfinder". Not for implementation slices of a known feature, milestone plans, general UI generation, standalone prototypes outside a decision map (use prototype), or adopting experimental code into production.

    Also in 3 projects that use this

  • AWSSkill

    map

    Map verified service capabilities against customer-provided control framework and threat model. Produces mapping-results.json, controls-catalog.md, and framework-mapping.md. Supports pluggable frameworks: CCMv4, NIST 800-53, CIS Benchmarks, ISO 27001, etc. Split into 4 focused sub-skills for better accuracy; this skill orchestrates them.

    Also in 3 projects that use this

  • dubwoahSkill

    ask

    Answer questions about the codebase without generating or modifying any code. Use when the user wants to understand how something works, find where something is defined, or explore the project structure. Invoked with /ask.

    Also in 3 projects that use this

  • javatarzSkill

    wiki

    Access and manage the local GitHub Wiki clone. Use this skill when you need to read wiki content (architecture, design docs, guides), update wiki documentation, check wiki status, or push wiki changes. The wiki contains high-level system design and architectural documentation.

    Also in 3 projects that use this

  • renatorozasSkill

    grill

    Adversarial code review. Don't let me ship until the changes pass your scrutiny.

    Also in 3 projects that use this

  • VercelSkill

    setup

    Set up Vercel CLI and project configuration. Use when the user says "set up Vercel", "configure Vercel", "link to Vercel", or "vercel init".

    Also in 7 projects that use this

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 44 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    It does not judge whether anything was big: No threshold is defined — see *The cutoff question* below, which is not optional.

    line 24
  2. Very specific:

    No interval given → use 30s: Do *not* use the parser's own 5-minute default here: five minutes means the command blocks for five minutes.

    line 38
  3. Very specific:

    Add --top N to change how many rows come back (default 10).

    line 52
  4. Very specific:

    If dist/bin/stockbit-live.js is missing from a clone, the source has not been compiled since it changed — npm run build there once.

    line 67
  5. Very specific:

    Read seconds, not the requested interval.

    line 123
  6. Distinctive:

    Place, cancel or modify an order.

    line 156

“Common” is measured: kept by at least 6 in 10 of the 44 edited versions of this file we could read. “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

Watch live IDX turnover

/watch <stocks|watchlist|all> <time-frame> <prompt>

/watch BBCA,ANTM 30s is anyone dumping
/watch watchlist 1m what is moving
/watch watchlist:Bandar 30s any unusual size
/watch all 5m where is the money going

What this actually does — read before answering

It takes two readings of the market some seconds apart and reports the difference. That is the whole mechanism. Δvalue / Δfrequency is the average rupiah size of the prints that landed in the window.

It does not watch continuously. One command, one window, one answer. Nothing runs afterwards, nothing will notify the user later. If they ask to be told when something happens, say plainly that this samples on demand and a background watcher does not exist yet.

It does not judge whether anything was big. No threshold is defined — see The cutoff question below, which is not optional.

Step 1 — read the arguments

They arrive as $ARGUMENTS (individually as $1, $2, …). If nothing was substituted, take them from the user's message after /watch.

positionmeaningaccepted
1 — scope, requiredwhat to look atBBCA · BBCA,ANTM · watchlist · watchlist:Name · all
2 — interval, optionalhow long a window to measure20s · 30s · 1m · 5m · realtime (also menit / detik)
3+ — prompt, optionalthe question, in the user's own words, either languageanything

Only the scope is required. Fill in the rest rather than refusing — a public user typing /watch BBCA should get an answer, not a usage error:

  • No interval given → use 30s. Do not use the parser's own 5-minute default here: five minutes means the command blocks for five minutes. Only use a long window when the user actually asked for one.
  • No prompt given → report what traded, ranked, and stop. No question to answer is not an error.
  • Nothing at all given → default to all 30s and describe the most active names. Do not ask a clarifying question first; run it, then mention the fuller form.

Watch for the common slip: the interval is positional, so a prompt with no interval shifts everything left. /watch BBCA is anyone dumping must be read as scope BBCA, interval 30s, prompt "is anyone dumping" — never as interval is. Only treat token 2 as the interval when it actually looks like one (a number with an optional unit, or a real-time word); otherwise it is the first word of the prompt.

Step 2 — run the sampler

stockbit-live scan <scope> <interval> --top 10

Set an explicit Bash timeout of the window plus 45 seconds. The default is two minutes, so anything from 2m up is killed mid-measurement with no output — which looks exactly like a failure of the market rather than of the timeout. A window over 8 minutes is refused by the CLI itself.

Add --top N to change how many rows come back (default 10). For all, more than ~15 rows is noise. Add --pretty only when a human will read the raw JSON.

If stockbit-live is not on PATH

Fall down this ladder — do not hardcode an install path, because the registered build gets repointed and a stale path in a markdown file is a silent misdiagnosis waiting to happen:

  1. Installed from npm, not globally linked:
    npx -y -p stockbit-mcp stockbit-live scan <scope> <interval> --top 10
  2. Running a local clone (the usual case when this server is registered as an MCP pointing at a dev build). Resolve the directory from the registration itself — mcpServers.stockbit.args[0] in ~/.claude.json is the one source of truth for which build is actually loaded — then:
    node dist/bin/stockbit-live.js scan <scope> <interval> --top 10
    On a machine that already has the helper, node "$HOME/.claude/helpers/stockbit-resolve.js" --path prints that directory.
  3. If dist/bin/stockbit-live.js is missing from a clone, the source has not been compiled since it changed — npm run build there once.

The other commands

scan measures turnover and ranks it. These do more.

signals — run the detectors

stockbit-live signals <scope> <interval> [prompt] --pretty

Takes two readings, runs every detector the prompt enabled, and puts the results through the alert engine (max 5 per interval, 25 per session, cooldowns, dedup). Reads order books only for symbols that actually traded, capped at 12 per pass — one request each.

Output carries alerts (what to show), suppressed (with the reason), marketWide (set when one signal tripped across most of the universe — show that line INSTEAD of the rows), and prompt.downgrades (requests that cannot be served as asked). Always show the downgrades.

explain <SYMBOL> — what actually printed

stockbit-live explain BRMS [HH:MM:SS] [HH:MM:SS]

Names the individual trades: time, price, lots, value, which side crossed the spread, and the broker on each side. Always lagged: true — the tape runs 8-10 minutes behind, so this explains something that already happened. Check brokersVisible: IDX closes broker codes during live trading, so identity may be withheld while the market is open.

brokers <SYMBOL> — who was on each side

stockbit-live brokers BRMS

End-of-day broker flow. Never an alert, always context. The useful column is lotsPerTrade (|netLots| / freq): a broker buying in many small pieces is retail-shaped flow; one buying the same lots in a few large trades is institution-shaped. Those boundaries are ours — IDX publishes none — and the output says so. If degraded is true, read degradedReason and do not interpret the labels.

Step 3 — read the output

JSON on stdout. ok: false always carries reason and detail; the process also exits non-zero, so neither a silent stdout nor a zero exit can be mistaken for a calm market.

fieldmeaning
deltas[]what traded, ranked by rupiah. value = rupiah in the window, trades = number of prints, averageTradeValue = rupiah per print, seconds = the real window length
confidencesingle — one print, so the average is that trade · few — 2–5 · averaged — more than 5, says nothing about any individual print
quiet[]in scope, present in both readings, traded nothing. A real observation.
unobserved[]in scope but absent from at least one reading, so there is no baseline. This is a gap in our data, not a quiet stock — never report it as one.
truncatedrows that crossed the --top cap. Mention the count if it is large.
symbolsSeenhow many symbols the reading covered, always ~100

Read seconds, not the requested interval. If a poll was slow the real window is longer, and describing a 40-second window as 30 seconds misstates every rate derived from it.

Failure reasons

reasonWhat to say
authThe Stockbit session is dead. Run /stockbit-status, then /stockbit-auth. Do not describe this as a market problem.
market-closedIDX is shut — give the WIB time and phase from detail. Offer --always, and say the readings will be identical.
session-resetThe counters restarted mid-window. Just run it again.
window-too-longOver 8 minutes cannot be measured in one call. Offer a shorter window.
bad-argumentsQuote detail — it names the offending token. Usually the prompt bled into the scope slot.
rate_limited / upstreamStockbit's side. Say so and offer to retry.

Step 4 — answer the prompt

Answer the user's actual question against the numbers. Give rupiah in a readable form (Rp 2.5 bn, Rp 39.7 M), name the window length, and say how many prints each figure covers.

The cutoff question — this is the part to get right.

If the prompt asks a yes/no about size — "is anyone dumping", "was that big", "ada transaksi besar?" — you do not have a defined answer. No threshold exists yet: it is an open decision waiting on measurement, and IDX mean print size varies roughly tenfold across the price range, so there is no safe constant to reach for.

So: report what the window measured and, if you offer a judgement at all, state the cutoff you used in rupiah. Never imply a settled rule exists. An unstated cutoff invented fresh each time is worse than a documented bad one — it cannot be reviewed, and it silently contaminates the very runs meant to decide what the real threshold should be.

confidence carries most of the honesty here. averaged over 300 prints says something about flow; it says nothing about any single trade. Do not turn it into "a large transaction went through".

What this cannot see

  • Individual prints in liquid names. The running-trade tape runs 8–10 minutes behind, so real-time detection is flow-level only. A single print is visible only when it dominates a short window in a thin name.
  • Broker identity. IDX closed broker codes in live running trade on 6 December 2021. Bandarmology is end-of-day — for that, use mcp__stockbit__broker_summary, and say it is T+1.
  • Anything outside the ~100 most active symbols. One request covers those. A stock with no turnover cannot print a large transaction, so the omission is mostly harmless — but say "the 100 most active", never "the whole market".

Never

  • Place, cancel or modify an order. This reads one public market endpoint; the trading tools are not part of this skill.
  • Report a failure as a quiet market.
  • Present unobserved symbols as having traded nothing.
  • State a big-transaction threshold as though it were established.

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.