stock-balance-sheet-review
Skill30 starsChanged 4 months agoMITView on GitHub
Review a US-listed company's balance sheet health for an equity-research workup. Covers net-debt/EBITDA leverage, current ratio, cash runway, goodwill concentration and impairment history, DSO trend, inventory days, off-balance-sheet items (operating leases, contingent liabilities), and pension underfunding. Trigger when analyzing financial-statement risk at L4 of the seven-layer X-ray framework, or when assessing whether the company can survive a downturn. Dispatched by stock-analysis-lead.
In this repository52 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- api-designSkill
- api-integration-testSkill
- constructive-thinkingSkill
- create-prSkill
- deep-researchSkill
- e2e-testSkill
- fuzzing-testSkill
- git-commitSkill
- go-benchmarkSkill
- go-ci-workflowSkill
- go-code-reviewerSkill
- go-concurrency-reviewSkill
- go-dependency-auditSkill
- go-error-reviewSkill
- go-logic-reviewSkill
- go-makefile-writerSkill
- go-observability-reviewSkill
- google-searchSkill
- go-performance-reviewSkill
- go-quality-reviewSkill
- go-review-leadSkill
- go-security-reviewSkill
- go-test-reviewSkill
- incident-postmortemSkill
- kafka-event-driven-designSkill
- load-testSkill
- local-transcriptSkill
- log-analyzerSkill
- mongo-migrationSkill
- monitoring-alertingSkill
- mysql-migrationSkill
- oceanbase-table-designSkill
- oracle-migrationSkill
- pg-migrationSkill
- readme-generatorSkill
- redis-cache-strategySkill
- security-reviewSkill
- stock-analysis-leadSkill
- stock-business-reviewSkill
- stock-earnings-quality-reviewSkill
- stock-industry-reviewSkill
- stock-management-reviewSkill
- stock-peer-comparison-reviewSkill
- systematic-debuggingSkill
- tdd-workflowSkill
- tech-doc-writerSkill
- thirdparty-api-integration-testSkill
- unit-testSkill
- update-docSkill
- writing-plansSkill
- yt-dlp-downloaderSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
Read the balance sheet from the most recent 10-Q (more recent than 10-K).
- Distinctive: line 34
Read the "Commitments and Contingencies" footnote in every review.
- Very specific: line 153
Every finding ID must start with BS — the prefix is the whole segment before the first hyphen.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Stock Balance Sheet Review
Purpose
Read the balance sheet for survivability. Most retail investors ignore the balance sheet until a recession exposes the leverage they didn't price in. Surface leverage and liquidity risks, goodwill bombs from past M&A, working-capital deterioration that signals customer-credit pressure, and off-balance-sheet hooks that compound in stress. Output a structured Findings set with quantitative thresholds.
When To Use
- Orchestrator dispatches balance-sheet review (always-on in Standard / Strict depth).
- User asks "is
<ticker>financially safe", "what's the debt situation", "can<ticker>survive a recession".
When NOT To Use
- Earnings quality / cash flow →
stock-earnings-quality-review - Business model →
stock-business-review - Capital allocation history →
stock-management-review - Competitive position →
stock-industry-review
Mandatory Gates
1) Execution Integrity Gate
Read the balance sheet from the most recent 10-Q (more recent than 10-K). If only 10-K is available, note the date staleness in Execution Status.
2) Industry-Norm Gate
Leverage tolerable for a utility is catastrophic for SaaS. Identify sector before applying thresholds. Default thresholds below are general; flag if your sector materially shifts them.
3) Off-Balance-Sheet Gate
Read the "Commitments and Contingencies" footnote in every review. Operating leases were brought on-balance-sheet by ASC 842 (2019), but contingent liabilities, guarantees, and purchase obligations are still footnote-only.
4) Goodwill-History Gate
A single goodwill impairment is a yellow flag; multiple is a red flag indicating chronic overpayment in M&A. Check 5-year impairment history, not just current year.
5) First-Hand Data Gate (mandatory)
Cite debt, cash, lease, pension, goodwill, AR and contract-liability figures from the manifest's first-hand financials.json (SEC EDGAR XBRL, via scripts/finlib/edgar.py) or directly from the filing — never from a search snippet. Any leverage ratio you state (Net Debt/EBITDA etc.) must use the named-input formulas in scripts/finlib/ratios.py so it survives the orchestrator's 口径 lint gate; in particular EV-based metrics require total_debt and cash_and_sti (a net-cash company has EV below market cap). Tag any second-hand figure as such and lower confidence accordingly.
Workflow
- Load 10-Q balance sheet and 10-K notes (Commitments, Goodwill, Pension).
- Run filing-pattern-gated scan.
- Compute ratios: Net Debt / EBITDA (using TTM EBITDA from financial-history), Current Ratio, Goodwill / Total Assets, DSO, Inventory Days.
- Compare to industry norms and to the company's 5-year history.
- Emit Findings.
Filing-Pattern-Gated Execution Protocol
Execution Order
- Locate balance sheet (10-Q latest), Goodwill note, Commitments note, Long-term debt note.
- For each checklist item, locate the specific line/note.
- HIT → compute ratio, compare to threshold, semantic-confirm.
- MISS → mark NOT FOUND.
- Report only Findings that breach thresholds OR show adverse trend.
- Include
Filing pre-scan: X/Y items hit, Z confirmed.
Balance Sheet Checklist (9 Items)
| ID | Item | Filing Section | Threshold |
| BS-01 | Net Debt / EBITDA | Long-term debt + Current portion − Cash; TTM EBITDA | Flag > 2× general; sector-aware (utility OK at 4×, SaaS bad at 2×) |
| BS-02 | Current Ratio | Current Assets / Current Liabilities | Flag < 1.0 (liquidity strain); < 1.5 in cyclical industries |
| BS-03 | Cash runway (unprofitable cos) | Cash / annual opex (or annual FCF burn) | Flag < 12 months — must refinance or issue equity |
| BS-04 | Goodwill / Total Assets | Balance sheet Goodwill / Total Assets | Flag > 30% — acquisition-driven; vulnerable to impairment |
| BS-05 | Goodwill impairment history | 5-year notes "Goodwill impairment" | Flag > 0 impairments in 5y as yellow; > 1 = red |
| BS-06 | DSO (Days Sales Outstanding) trend | AR / (Annual revenue / 365), 3-year trend | Flag rising > 10 days YoY without explanation in MD&A |
| BS-07 | Inventory days trend (non-tech) | Inventory / (Annual COGS / 365) | Flag rising > 15% YoY (slowing sell-through / obsolete inventory risk) |
| BS-08 | Off-balance-sheet items | "Commitments and Contingencies" note | Flag material guarantees, purchase obligations, or contingent liabilities |
| BS-09 | Pension underfunding | Pension footnote (defined-benefit only) | Flag underfunded amount > 5% of market cap |
Severity Rubric
- High: Survivability-threatening. Net Debt/EBITDA > 4× in cyclical sector; cash runway < 6 months; goodwill > 50% of assets; multiple recent impairments.
- Medium: Adverse but manageable. Leverage rising; DSO trending up; contingent guarantee material but defined.
- Low: Watch-list. Inventory days slightly elevated; pension modestly underfunded.
Evidence Rules
- Compute the ratio yourself; cite the source lines.
- Include the threshold you're comparing against (general or sector-specific).
- Trends: cite at least 3 years of values, not a single snapshot.
Output Format
Findings
[High|Medium|Low] Short Title
- ID:
BS-NN - Citation: balance sheet line / note, with periods
- Evidence: computed ratio + threshold + trend
- Implication: what this means under stress
Suppressed Items
Execution Status
Filings reviewed: 10-Q (Q3 2025), 10-K (FY2024) for footnotes
Filing pre-scan: 9/9 items hit, 2 confirmed as findings
Industry norm applied: <sector>Summary
One line: N High / M Medium / K Low — most material: <BS-NN short title>.
Machine-Readable Findings Block (mandatory)
End the reply with exactly one fenced block tagged findings-json, carrying
Worker Findings Contract v1 (full schema and error codes:
stock-analysis-lead/references/worker-contract.md). The orchestrator synthesizes
the verdict from this block only — anything stated in the Markdown above but
omitted here does not reach the report. Everything above the fence is for the
human reader.
{
"contract_version": "1",
"worker": "stock-balance-sheet-reviewer",
"prefix": "BS",
"status": "OK",
"depth_mode": "<echo the dispatched depth>",
"archetype_applied": "<echo the dispatched archetype>",
"archetype_challenge": null,
"findings": [
{
"id": "BS-NN",
"severity": "High|Medium|Low",
"title": "<= 80 chars",
"citation": {"source": "10-K", "locator": "<item/page/note>", "fiscal_period": "FY2025"},
"evidence": "direct quote <= 60 words, or a computed figure with its inputs",
"implication": "one sentence on what this means for the thesis",
"confidence": "first-hand|second-hand"
}
],
"positives": [],
"data_gaps": [],
"checklist_coverage": {"items_total": 9, "items_checked": 0, "items_not_found": 0, "ids_not_checked": []},
"mandatory_checks_run": []
}Contract rules that fail validation if broken:
statusis one ofOK/DEGRADED/SKIPPED/REFUSED; any value other thanOKrequires astatus_reason. The gate returnsSKIPPED (...)in prose and"status": "SKIPPED"here.- Every finding ID must start with
BS— the prefix is the whole segment before the first hyphen. citationis an object, never a bare string;locatorandfiscal_periodmust be non-empty. A finding with no real citation is suppressed, not emitted.source: "aggregator"forcesconfidence: "second-hand"— this is what makes the first-hand data rule checkable rather than aspirational.checklist_coverage:items_checked + items_not_foundmust reachitems_total(9), so "the checklist ran to completion" is verifiable.archetype_challenge:nullwhen the dispatched archetype fits. When the evidence says it does not, file{"proposed", "reason", "evidence"}instead of silently analyzing against thresholds you believe are wrong — this is the only sanctioned way to disagree with the orchestrator's classification.mandatory_checks_run: list the archetype-specific check IDs the dispatch marked REQUIRED. Omitting one that was required fails validation.
Self-check before replying:
python3 <path-to>/stock-analysis-lead/scripts/finlib/worker_contract.py \
validate --reply <this-reply>.md --expect-worker stock-balance-sheet-reviewerNo-Finding Case
No balance-sheet findings — leverage moderate, liquidity ample, no goodwill concentration.
Notable positives: Net Debt / EBITDA = 0.4×; Cash > Total Debt.Load References Selectively
references/balance-sheet-red-flags.md— load when assessing goodwill concentration, off-balance-sheet items, or pension liabilities; contains the sector-norm leverage table, historical examples of goodwill bombs, and the Commitments-and-Contingencies decoding guide.
Review Discipline
Balance sheet failures are slow until they're sudden. A leveraged company looks fine in a bull market and dies in a recession. Be unforgiving on High thresholds; the orchestrator will weight these heavily in the Bear scenario.
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.

