| name | financial-output-qa-gate |
| description | Use when financial, investment, trading, valuation, accounting, market-data, broker-journal, or client-facing outputs need a final QA gate before delivery or persistence. This skill checks source logs, timestamps, missing data, Not investment advice, forbidden target-price/rating/position-sizing/return-promise language, unconditional trade instructions, artifact references, and output-contract compliance without performing new analysis or inventing data. |
Financial Output QA Gate
Use this skill after the domain workflow has produced a draft output and before
the result is delivered as a productized financial artifact or stored under
.research/.
This skill is a checker, not an analyst. Do not add new market views, valuation
judgments, recommendations, or trade ideas while using it.
Required Inputs
- Draft output or machine-readable record.
- Expected workflow recipe from
.agents/workflows/ when available.
- Shared contract from
.agents/references/output-contract.md.
- Data registry from
.agents/references/data-capability-registry.md when
market data is used.
- Display profile from
.agents/references/display-profiles.md when the output
will be shown in App or CLI.
Checks
source_log exists and records source name, endpoint/interface, parameters,
retrieval time, status, and missing fields.- Market data includes trading date, retrieval timestamp, unit/currency, and
adjustment basis when relevant.
qa_status exists and records pass/warn/fail checks.Not investment advice is present for investment, trading, valuation,
portfolio, credit, or client-facing conclusions.- Missing data and source gaps are explicit.
- Artifact references exist when the workflow writes
.research/ outputs.
- The output does not contain target prices, buy/sell ratings, personal
position sizing, return promises, or unconditional buy/sell instructions.
- Technical analysis, factors, or backtests are not presented as a standalone
reason to trade.
- App or CLI display uses the selected display profile without hiding material
source gaps or failed QA checks.
Tooling
Use tools/check_research_integrity.py when a JSON record is available:
python3 tools/check_research_integrity.py --input path/to/record.json --profile a_share_decision
python3 tools/check_research_integrity.py --input path/to/record.json --profile general_finance
python3 tools/check_research_integrity.py --input path/to/record.json --profile journal_shadow
Output
Return a compact QA result:
status: pass, warn, or fail.blocking_issues: issues that must be fixed before delivery.warnings: non-blocking limitations.checks_performed: checklist summary.source_gaps: unresolved data gaps.guardrail_status: whether unsupported financial outputs were blocked.
