networth
Skill0 starsChanged 54 days agoMITView on GitHub
Live whole-portfolio snapshot, all accounts. Use when asked what the portfolio is worth right now, to snapshot the accounts, whether concentration sits near doctrine ceilings, or how today's move changed theme-alpha thesis exposure. Broker-authoritative holdings (the broker read MCP) + live quotes -> value, allocation, thesis exposure, concentration vs ceilings, dated snapshot. Not a single-ticker quote or a past-date lookup.
- Reads credentials
In this repository18 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 43
If any price fetch fails, label that position PRICE UNAVAILABLE and exclude from totals (never silently omit.
- Very specific: line 49
Read relevant thesis/account doctrine from permitted references and wiki/hot.md pending items.
- Very specific: line 52
Read Atlas/sources/investing/ref-portfolio-doctrine.md: the thesis map (ticker -> thesis) and the machine doctrine: block ceilings -- cite …
- Distinctive: line 87
Every held ticker accounted for: in the positions table, or labeled PRICE UNAVAILABLE and excluded from totals.
- Distinctive: line 88
All reported asset quantities are broker-authoritative.
- Distinctive: line 91
No hardcoded roster used anywhere (cross-account set, thesis rows, caveats all derived this run).
≈ “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
Financial evidence contract: docs/financial-analysis-contract.md governs source
quality, temporal/basis semantics and missing-data handling across harnesses.
For material numeric batches, run python tools/fis/evidence.py <envelope.json>
before output, or disclose manual review and any unverified semantics.
Plugin interoperability: .agents/skills/finance-data/SKILL.md owns typed
evidence intake and portable Finance/Data handoffs. This skill retains its
financial procedure and gates. Probe callable tools at run time; installed
plugin metadata alone does not establish a usable connector or source schema.
Mode routing (Pattern 6 deterministic; invocation modes)
| Syntax | Behavior |
/networth | Full snapshot: MCP holdings + price-fetcher dispatch + calc + dated snapshot write + peripherals, atomic commit |
/networth --preview | All phases in memory; render the would-be snapshot to stdout; SKIP all writes; F11 never set |
Quality Rules
- ZERO portfolio data in this file. Holdings, share counts, cross-account sets, thesis membership, caveats, reserves: ALL derived at run time from their owning sources. A ticker list or dollar figure appearing in this SKILL.md is a defect (the stale-literal failure mode), not a convenience.
- Dollar amounts from exact share counts x exact prices. Never estimate or round share counts.
- Share counts, quantities and basis require runtime broker read evidence for every account and asset class. Missing coverage is UNVERIFIED; report known subtotal and missing scope, never private-file fallbacks or zero estimates.
- Doctrine threshold math (concentration ceilings, thesis weights) anchors to regular_market_close ONLY -- extended-hours prices are reported but NEVER trigger doctrine evaluations (ref-portfolio-doctrine law).
- Any position whose caveat line (interpretation layer) flags unreliable cost basis gets NO P&L line -- current-value-only, with the caveat named.
- Subagent dispatch is MANDATORY where the harness provides one (Phase 1 dispatches
price-fetcherwhen an Agent tool exists; that subagent is opus-pinned by its own definition -- never above the opus subagent ceiling, cost directive 2026-07-09). Pre-emptive skip ("script will probably fail", "I'll just WebSearch") is FORBIDDEN; the legitimate fallback fires ONLY on contract violation, actual dispatch failure, or a harness with no Agent tool; pre-emptive skip surfaces in Phase 4 as DEVIATION. - If any price fetch fails, label that position
PRICE UNAVAILABLEand exclude from totals (never silently omit; never estimate). - ASCII only in the snapshot + every appended line (Pattern 22).
Phase 0 -- Broker-authoritative holdings (MCP first)
- Load the broker read tools (ToolSearch on harnesses that defer tool schemas); call
get_equity_positions(ALL accounts configured for the deployment) +get_portfolio(account totals, cash, crypto aggregate). This is the SOURCE OF TRUTH for tickers, share counts, cost basis, account mapping. - Read relevant thesis/account doctrine from permitted references and wiki/hot.md pending items. Never read
.raw/,private/,finance/,credentials/,.env*,auth.json, or*.local.md; obtain current account facts through broker read tools. - DERIVE (never hardcode): CROSS_ACCOUNT_SET from current-session broker accounts; CAVEAT_MAP from supported broker-basis/corporate-action evidence and session constraints. RESERVES are explicitly dated user earmark decisions from permitted sources, separate from broker-verified cash. A stale or missing earmark is unresolved, never zero or an inferred account balance. 3b. CRYPTO COUNTS: discover current tool capability; never assume availability from an old tool list. Per-coin quantities require a successful authorized broker read. If only a broker aggregate is available, preserve that aggregate without inventing per-coin allocation. If neither exists, crypto and whole-book concentration are UNVERIFIED.
- Read
Atlas/sources/investing/ref-portfolio-doctrine.md: the thesis map (ticker -> thesis) and the machinedoctrine:block ceilings -- cite by path, never restate from memory. - Availability guard: missing broker coverage -> quantities and affected account/whole-book totals UNVERIFIED. No private-file fallback. Continue only with a clearly scoped verified subtotal.
Phase 1 -- Live prices (price-fetcher dispatch; MANDATORY where available)
Dispatch price-fetcher when the harness exposes an Agent tool; otherwise execute the documented inline fallback below verbatim. Input: {equities: [<derived list>], crypto: [<derived list>]}. The subagent runs broker-authoritative equity quotes (Tier 0) then tools/fetch-prices.py (Tier 1) then per-ticker retry (Tier 2), and returns a structured JSON quotes map. With no Agent tool, the main loop runs those same three tiers inline in the same order and assembles the same quotes map -- the tier order, the validation below, and the C5 gate are identical either way.
Validate the return: timestamp; quotes map with every held ticker present in quotes OR failures (no silent omissions); per-quote market_session (pre-market | regular | after-hours | closed) + source + broker_authoritative + extended-hours fields; caveat fields honored per CAVEAT_MAP (e.g. a no-P&L basis flag rides through as caveat).
C5 broker-authoritative gate (red-team 2026-07-09; enforces the price-fetcher parent contract): during a REGULAR session with the broker MCP present, assert mcp_price_count >= held-equity count. Shortfall -> ONE re-dispatch (or one inline retry where no Agent tool exists); second shortfall -> HALT the portfolio-math phases (do NOT proceed with non-broker prices for regular-session portfolio math; report which tickers fell through). Structural MCP absence (harness without the broker server, headless, server down) is the degrade path, not a HALT: proceed on script prices with a price-unconfirmed flag in the snapshot frontmatter + Notes. WebSearch prices for EQUITIES during a regular session are FORBIDDEN (the fallback price domains are on the AGENTS.md blocklist and defeat the broker-authoritative premise); the inline WebSearch fallback below is legal only outside regular session or for assets the script cannot cover.
Contract violation or dispatch failure -> ONE re-dispatch -> inline WebSearch fallback WITHIN the C5 bounds above: batch the DERIVED holdings list (equities, ETFs, crypto -- groups built from Phase 0, never a hardcoded roster), record price + freshness per ticker. Record the dispatch tally for Phase 4.
Phase 2 -- Calculate
Per-ticker live-valuation price source: market_session in (after-hours, pre-market) AND non-null extended_hours_last -> use it, surfaced as "AH-priced"/"PM-priced"; else regular close. Crypto always price (24/7). Closed market -> last regular close.
Doctrine anchoring: concentration + thesis-weight aggregation uses regular_market_close values ONLY. Extended-hours valuation is reported but never evaluated against the doctrine ceilings.
Per position: current value = shares x live price; gain/loss + % vs cost basis. HARD P&L RULE (independent of CAVEAT_MAP parse success): any transferred-in position whose basis integrity cannot be POSITIVELY established gets NO P&L line -- current value only, reason stated (CAVEAT_MAP flags are one establishment path, absence of a clean broker basis is itself disqualifying). CRYPTO RECONCILIATION: sum of broker-verified per-coin values MUST be reconciled against the get_portfolio broker crypto aggregate; divergence beyond price-movement tolerance -> flag line in Notes ("per-coin quantities may be stale; broker aggregate differs by X%"). Aggregate: per-account value (positions, crypto and cash as each account reports them), combined; cash; allocation by thesis (DERIVED map; regular-close anchored) and by account; top-5 by weight; concentration (top single name %, top-5 %, crypto %) vs the doctrine ceilings read in Phase 0 -- state headroom or breach per ceiling; extended-hours delta ($ + portfolio % shift) when any held position has non-null AH change.
Thesis attribution -- DUAL BASIS (red-team 2026-07-09): the doctrine thesis map has OVERLAPPING memberships (some tickers belong to more than one thesis) and the doctrine: block's concentration.basis: effective does not define the shared-ticker split -- the operative half-weight convention lives only in snapshot precedent. Until the rule is ratified into the block: compute and report BOTH bases against each ceiling -- STRICT (shared tickers full-weight in every thesis they belong to) and EFFECTIVE (per the most recent prior snapshot's stated convention, cited by path) -- and NEVER silently pick one; where the two straddle a ceiling, say so explicitly. Standing FOLLOWUP each run until closed: /decide ratify-thesis-attribution-rule -- encode the shared-ticker split into the doctrine block (re-fingerprint + ratification; Atlas is human-write-only).
Phase 2.5 -- Historical comparison
Latest prior wiki/investing/snapshots/networth-*.md: change since ($ and %), best/worst performer, thesis-allocation drift vs targets, reserve/earmark progress -- compare the relevant account cash against the CURRENT reserve + earmark lines derived in Phase 0.3 (never against a remembered figure).
Phase 3 -- Compose snapshot
Writes to wiki/investing/snapshots/networth-<YYYY-MM-DD>.md (plus the Phase 4 peripherals: wiki/hot.md bump, daily note, sessions-log); same-day collision -> -HHMM variant (archival rule). Frontmatter: canonical + type: snapshot + valuation_session: <regular|after-hours|pre-market|closed> + related: investing-moc.
Body: # Portfolio Snapshot -- <date>; combined value + prices-as-of timestamp + source mix (broker-authoritative count / script / fallback); ## By Account table; ## All Positions table (Ticker | Account | Shares | Price | Value | Weight | Gain/Loss -- CROSS_ACCOUNT_SET rows shown combined with per-account sub-rows; caveat-flagged rows show current value only); ## By Thesis table (rows = the DERIVED thesis map; dual-basis: Strict Value/Weight + Effective Value/Weight | Positions); ## Concentration (top name %, top-5 %, crypto % -- each vs its doctrine ceiling with headroom, both bases where they differ); ## Change vs prior snapshots (Phase 2.5 output: delta $/%, best/worst performer, drift, reserve progress); ## Notes (every applicable CAVEAT_MAP line + reserve status + crypto-reconciliation result + UNVERIFIED coverage / price-unconfirmed disclosures). Every quantitative claim carries prov: (mcp:* > script:* > file:* > web:*).
Phase 3.5 -- Pre-Output HALT gate
- Every held ticker accounted for: in the positions table, or labeled PRICE UNAVAILABLE and excluded from totals.
- All reported asset quantities are broker-authoritative; missing asset/account coverage is UNVERIFIED and excluded from unsupported whole-book claims. Any aggregate/per-name mismatch is disclosed. Price/session provenance applies to every harness.
- Doctrine math regular-close anchored; ceilings cited by path; headroom/breach stated per ceiling on BOTH attribution bases; straddles called out.
- Transferred-in positions without established basis integrity have no P&L line (hard rule, CAVEAT_MAP-independent); caveat named in Notes.
- No hardcoded roster used anywhere (cross-account set, thesis rows, caveats all derived this run).
- ASCII clean; -HHMM collision resolved; no write targets
.raw/ private/ finance/ credentials/ Atlas/. - F11 set before writes (
--preview: no writes, F11 never set). - Sessions-log entry + daily-note line composed; hot.md
last_networthbump planned.
ANY failure -> HALT with the item number; F11 stays on.
Phase 4 -- Atomic writes + peripherals
- Set F11 (
touch .claude/state/auto-commit-disabledwhere the per-write auto-commit hook runs; harnesses without that hook honor the same lifecycle by instruction -- [[ref-execution-discipline]]); write the snapshot. - hot.md frontmatter bump:
last_networth: <today>(field exists in hot-md-v2 schema; update only, sha256-preserve elsewhere). - Daily note
## Observationsline:- [HH:MM] Portfolio snapshot: $<combined> (<valuation_session>; <source mix>); sessions-log entry per canonical schema. - Dispatch report (one of):
Phase 1 price-fetcher: DISPATCHED|INLINE (no Agent tool in this harness)|FALLBACK (<reason>)|DEVIATION (<judgment>)-- DEVIATION is a discipline breach, logged to sessions-log. - F14 narrow stage (snapshot + hot.md + daily + sessions-log); commit
agent: networth <date> -- <combined> (<session>); F17 verify; F11 clear.
Failure modes
| Symptom | Likely cause | Fix |
| MCP tools absent | Read connector absent, denied, or unavailable | Report UNVERIFIED; no private-file fallback; repeat when runtime evidence becomes available |
| price-fetcher return missing market_session | Contract violation | One re-dispatch -> inline WebSearch fallback; tally in Phase 4 |
| A ticker in neither quotes nor failures | Silent omission (contract breach) | Treat as violation -> re-dispatch; never fill from memory |
| Concentration flips a ceiling on AH prices | Doctrine anchoring broken | Recompute on regular close; AH is report-only by law |
| Same-day collision | Prior snapshot today | -HHMM variant (archival rule) |
Coordination
- /brief -- reads the latest snapshot for portfolio movers; flags >5% hot.md baseline drift as this skill's trigger; /networth owns the portfolio-refresh scope (retired from /brief Phase 15).
- /invest -- Phase J portfolio-fit math shares the same MCP-first counts law; snapshots are its exposure baseline.
- /gate + pretrade staircase -- concentration numbers here are advisory context; execution-time ceiling checks live in tools/pretrade_gate.py against the trusted broker book (never this snapshot).
- /retro -- sessions-log entry consumed at session end.
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.

