Qualitative Filtering

Methodology fused from professional trading and investment frameworks; all text is an original paraphrase.

Defaults

Preflight

Run canonical pre-flight per contracts/preflight.md. Include the X-Agentii-Trace header on every tool call per contracts/x-agentii-trace-header.md — carry the _run_id from your first tool result and name yourself (and your parent, if you were spawned).

Data Source Priority

1. Qualitative methodology — references/qual-methodology.md (bundled MOP-KPI-Catalyst framework)
2. Company disclosures — SEC filings (Business Description, Risk Factors, MD&A) via agentii MCP
3. Earnings transcripts — search_documents(ticker={T}, form_type="earnings_call_transcript") → read_source_outline → read_source_pages (citation prefix ect<N>; pages carry section_type in session_title and guidance/forward_looking/analyst_questions in labels)
4. Strategy and case knowledge — search_investment_strategies + search_investment_cases + search_by_analogue

Methodology

Retrieval Scope

unstructured_document_search (earnings call transcripts + SEC disclosures)

Retrieval Strategy

Ownership & insider signals: search_institutional_holdings (top-10 holders + whale portfolios, direction=accumulating|reducing|new|exited) and search_insider_trades (Form-4 transactions with SEC URLs) are available as signal inputs.

Three-layer protocol from contracts/retrieval.md: the qualitative framework is bundled in references/qual-methodology.md. Earnings transcripts via search_documents(form_type="earnings_call_transcript") → read_source_pages (Layer 1→3). Strategy frameworks and historical analogues via MCP knowledge tools. Detailed methodology and catalyst classification in references/qual-methodology.md.

Temporal Scope

See frontmatter temporal_scope block.

Tool Allowlist

See frontmatter allowed_tools.

Protocol

This skill implements qualitative investment analysis through a three-stage framework: MOP → KPI → Results. Companies do not publish an "MOP" — the analyst infers it by identifying KPIs first, then reverse-engineering the strategic plan. The analyst verifies the chain is intact and credible. A broken chain is the highest-quality short signal.

Detailed methodology: management assessment framework (Alpha/Beta/Delta), board quality checklist, mosaic theory triangulation, catalyst taxonomy, sector-specific KPI templates, and the 20+ pattern red flag catalog are in references/qual-methodology.md.

Critical distinction: Good company ≠ good stock. For the 20-60 day horizon, catalysts are required.

Steps

1. KPI Identification: Identify industry-specific KPIs per the reference templates (SaaS, retail, manufacturing, financial services, healthcare). Determine leading vs. lagging. Assess consistency (changing KPIs = red flag), auditability, relevance. Map trends over 8+ quarters. KPI divergence from sector norms often explains quantitative outlier signals.

2. MOP Analysis (5-dimension scorecard, each 0-10, composite < 25 = high risk): Extract from earnings calls, presentations, MD&A. Infer the MOP: forward-looking statements → recurring themes → strategic narrative → test consistency/credibility. Score on: Track Record (30%, 3yr+ guidance vs. actuals), Consistency (20%), Realism (20%), Alignment (15%, insider ownership + compensation structure), Disclosure Quality (15%). Red flags: transformational M&A without plan, repeated guidance misses, high SBC with low hurdles, C-suite turnover within 18 months.

3. Management Team (Alpha/Beta/Delta): Alpha (CEO) — track record, capital allocation, communication style. Primary Betas (CFO, CPO, CTO, Corp Dev, CMO) — depth and tenure. Red flags: cluster departures, CFO departure near guidance, cluster insider selling.

4. Board Assessment: Independence ≥ 75%, expertise present, financial expert on audit committee. "Political incest" check: management/board overlap. Red flags: classified board, supermajority voting, tenure > 15yr, CEO as Chair, related-party transactions.

5. Industry Analysis: Five Forces and SWOT as thinking prompts, not rigid boxes. Read competitor 10-Ks to cross-check management narrative. Do not trust management pronouncements on competition — verify independently. "Explain to 10-year-old" test: describe in 1-2 sentences. "3-5 factors" rule: identify drivers that matter; more than 5 = spread too thin. For formal peer-set construction and relative benchmarking, defer to the peer-bench skill rather than rebuilding it here.

6. Consensus Reconstruction (run after steps 1-5, never before — reading consensus early anchors the analysis to the expectation it is meant to test): Establish the published sell-side average and its dispersion, then triangulate the effective buy-side expectation, which typically moves ahead of the published figure. Weight recent revisions over the stale average. Output a range with a direction, never a point estimate. Wide dispersion = no consensus exists, so the disconnect framing does not apply; tight dispersion with stale revisions is the highest-value setup. If the buy-side bar cannot be triangulated, mark the disconnect unquantified and flag a coverage gap rather than substituting the published number. State the variant view as: market expects X, evidence indicates Y, because [KPI/MOP finding], closing when [catalyst] by [date].

7. Catalyst Identification: Identify all catalysts within 20-60 days. Classify: Earnings / Corporate Action / Regulatory / Management / Industry / Macro. Assess: specificity (dateable?), magnitude (≥ 15% high, 5-15% standard, < 5% insufficient), probability, binary vs. spectrum (binary → reduce size). Tumbleweed test: < 1 non-earnings press release/month = avoid. Catalyst stacking: multiple = higher conviction; zero = investment, not trade.

8. Red Flag Scan: Scan against catalog (see reference). 3+ flags = hard stop for longs. Key flags: non-recurring charges in 3+ of 4 quarters, SBC > 10% revenue, GAAP losses + non-GAAP profits, trade data contradicts management, competitor filings describe different dynamics.

9. MCP Integration: search_investment_strategies(kind=qualitative) → search_investment_cases(domain=catalyst_driven) → search_by_analogue(event_type, company_situation). Handoff: conviction score (1-10), quantified consensus disconnect (or unquantified), ranked catalyst calendar, KPI summary, MOP score, management/board flags, red flag count.