earnings-preview-med
Med-adapted earnings preview: consensus estimates, historical surprises, guidance sensitivities, and the FDA-catalyst overlay (PDUFA/AdCom/trial readouts near the print) for biotech/pharma names.
Install
npx skills add https://github.com/agentii-ai/agentii-investment-intelligence/tree/main/plugins/vertical-plugins/bio-pharm/skills/agentii/earnings-preview-med
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install agentii-ai-agentii-investment-intelligence@llmmart
git clone https://github.com/agentii-ai/agentii-investment-intelligence.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole agentii-ai/agentii-investment-intelligence collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Methodology inspired by publicly taught earnings-preview frameworks; all text is an original paraphrase.
Defaults
| Parameter | Default Value | Rationale |
|---|---|---|
| surprise_window | 8 quarters | Standard surprise history window |
| include_catalysts | true | Med prints move on catalysts as much as EPS |
| guidance_sensitivity | true | Guidance is the swing factor for pharma |
Preflight
Run canonical pre-flight per contracts/preflight.md. Propagate X-Agentii-Trace per contracts/x-agentii-trace-header.md.
Triggers
- "Preview [biotech ticker]'s upcoming earnings."
- "What should I expect at [ticker]'s next print?"
- "How has [ticker] surprised historically?"
- "Which catalysts land near [ticker]'s earnings date?"
- "Build an earnings preview with consensus estimates."
- "What's the guidance risk for [ticker] this quarter?"
- "Summarize the last few quarters for [ticker]."
- "Earnings + FDA calendar overlap for [ticker]."
- "What are the swing factors for [ticker]'s print?"
- "Historical reaction to [ticker]'s earnings surprises."
- "What changed in my estimates, thesis, and positioning ahead of [ticker]'s print?"
Production Grounding
- Med prints have TWO drivers: financials (revenue/EPS/guidance) and catalysts (PDUFA/AdCom/readouts). The catalyst overlay is mandatory — a clean quarter can be undone by a CRL days earlier.
- Every preview closes with model-vs-consensus deltas (our modeled numbers vs consensus from
search_earnings_calendar, signed with the driver named) and the three what's-changed vectors (estimates / thesis / positioning) — estimates move first, thesis and positioning follow only when the facts justify them. - For pre-revenue biotechs, the print is mostly about cash runway + pipeline updates; consensus EPS is secondary.
- Grounding frameworks:
references/knowledge-frameworks.md(道/法 review knowledge).
Data Source Priority
search_earnings_calendar— estimates, actuals, surprise history, next date.search_xbrl_facts— revenue/EPS/margin trends.search_documents/read_source_*— prior-quarter commentary and guidance.- Knowledge layer:
search_investment_casesfor historical print reactions.
Methodology
Retrieval Scope
unstructured_document_search
Retrieval Strategy
- Resolve the earnings event:
search_earnings_calendarfor dates/estimates/surprises (consensus values). - Pull fundamentals trend:
search_xbrl_factskey line items (our modeled numbers). - Compute model-vs-consensus deltas on each key line item; state assumptions; annotate coverage_gap where no consensus value exists.
- Slot-refresh the three what's-changed vectors: estimates / thesis / positioning.
- Catalyst overlay: nearest PDUFA/AdCom/readout vs print date.
- Ground with historical cases (print reactions) via knowledge tools.
Temporal Scope
See frontmatter temporal_scope block.
Tool Allowlist
See frontmatter allowed_tools.
Protocol
- Event & estimates
- Fundamental trend
- Model-vs-consensus deltas
- What's-changed vectors (estimates / thesis / positioning)
- Catalyst overlay
- Swing-factor synthesis
Modes
- Standard (default): estimates + surprises + guidance.
- Pre-revenue: runway + pipeline + readout framing.
- Catalyst-overlap: print framed around nearby FDA events.
Tool Fallbacks
| Failure | Fallback |
|---|---|
| search_earnings_calendar empty | Use filings (search_documents) for dates; annotate coverage_gap |
| No catalyst data | Flag "catalyst overlay unavailable" |
| Knowledge tools empty | Proceed with structured data only |
Output File
{ticker}/{YYYY-MM-DD_HHMM}_earnings-preview-med_{affix}.md
Output Structure
- Executive Summary — setup for the print in 2-3 sentences
- Consensus & Surprise History — estimates table + surprise record
- Model vs Consensus — our modeled numbers vs consensus (
search_earnings_calendar) with signed deltas and named drivers - What's Changed — the three vectors: estimates / thesis / positioning (estimates move first, ratings last)
- Guidance & Swing Factors — guidance risk analysis
- Catalyst Overlay — FDA events near the print
- Historical Context — cases with /v/ citations
- Coverage Gaps — degraded flags
Error Handling
| Error | Fallback |
|---|---|
| Estimates missing | Present fundamentals trend only; flag |
| Date uncertain | Use calendar's best estimate + annotate |
Memory Load
See contracts/memory-load.md.
Snapshot
See contracts/snapshot-synthesis.md.
Final Summary (TUI)
Include ### Key Citations block with 0-10 clickable /v/ URLs.
References
contracts/citation-and-memory.mdcontracts/output-frontmatter-schema.mdcontracts/memory-load.mdcontracts/snapshot-synthesis.mdcontracts/preflight.mdreferences/knowledge-frameworks.md
Files (agentii-investment-intelligence)
-
references
-
knowledge-frameworks.md 2.2 KB
# Med Knowledge Frameworks (shared) Full regulatory grounding lives in the fda-catalyst-analysis skill's reference: `../../fda-catalyst-analysis/references/knowledge-frameworks.md` (FDA/EMA process, AdCom mechanics, six scrutiny axes, catalyst sizing, med metrics). ## 道/法 layered review knowledge (runtime retrieval) Retrieve at runtime via `search_investment_strategies(sectors=med, layer_tags=L2)` and cite with /v/ URLs: - 道 (L1): Substantial Evidence Standard · Totality of Evidence & Benefit-Risk Balance · Safety Signal Characterization · Patient-Centric Risk-Benefit Context · Clinical Meaningfulness of Endpoints - 法/器 (L2): Comparator Selection Adequacy · Missing Data Sensitivity · Subgroup Consistency · Surrogate Endpoint Validation · RWE Credibility · REMS Effectiveness · Trial Design Integrity · Vote Tally Interpretation Matrix · Red Flag Screening Checklist ## Structured AdCom calendar `search_adcom_meetings` (committee/product/ticker/date/vote filters, 122 meetings, 113 with votes) + `get_adcom_meeting` (briefing-doc inventory). ## What's-changed vectors & model-vs-consensus deltas (earnings preview) - **Three what's-changed vectors**: every preview closes by slot-refreshing three vectors — estimates (what numbers moved and why), thesis (what the quarter does to the investment case), positioning (where investor attention sits and the per-name bull/bear pivot conditions). Slot-refresh order: estimates move first; thesis and positioning follow only when the facts justify them. - **Model-vs-consensus deltas**: each key line item states our modeled number against the consensus value pulled from `search_earnings_calendar`, with the delta signed and the driver named. Where no consensus value is retrievable, the delta degrades to a `coverage_gap` with the required-input list — never a fabricated comparison. - **Uncertainty discipline**: scenario-flexing stays at market level, POS at asset level (peak × POS, changes logged with reasons); explicit abstention where a swing factor is unquantifiable. - **Buy-side lens**: client-question framing ("one common question we have received"); per-name bull/bear pivot conditions stated, not implied. -
modes.md 1.2 KB
# earnings-preview-med — Analyst Mode Definitions Derived from the skill's own methodology structure (scripts/mode_backfill.py, spec 046 M1). ### Mode: defaults **Objective**: Defaults analysis per the skill's methodology (see SKILL.md sections). ### Mode: preflight **Objective**: Preflight analysis per the skill's methodology (see SKILL.md sections). ### Mode: triggers **Objective**: Triggers analysis per the skill's methodology (see SKILL.md sections). ### Mode: production-grounding **Objective**: Production Grounding analysis per the skill's methodology (see SKILL.md sections). ### Mode: data-source-priority **Objective**: Data Source Priority analysis per the skill's methodology (see SKILL.md sections). ### Mode: drug **Objective**: Drug modality analysis per the skill's methodology (see SKILL.md sections). ### Mode: biologic **Objective**: Biologic modality analysis per the skill's methodology (see SKILL.md sections). ### Mode: device **Objective**: Device modality analysis per the skill's methodology (see SKILL.md sections). ### Mode: vaccine **Objective**: Vaccine modality analysis per the skill's methodology (see SKILL.md sections).
-
-
SKILL.md 5.7 KB
--- name: earnings-preview-med description: "Med-adapted earnings preview: consensus estimates, historical surprises, guidance sensitivities, and the FDA-catalyst overlay (PDUFA/AdCom/trial readouts near the print) for biotech/pharma names." sectors: [med.medicines_biotech, med.medical_devices] multi_ticker_semantics: single_target temporal_scope: default_quarters: 4 max_quarters: 8 description: "Preview window default 4 quarters; up to 8 for guidance trajectories." allowed_tools: - search_earnings_calendar - search_xbrl_facts - get_company_profile - search_documents - read_source_outline - read_source_pages - search_investment_cases - get_investment_case - search_investment_strategies - get_investment_strategy - search_by_analogue - search_knowledge_entries - get_knowledge_entry retrieval_scope: unstructured_document_search min_tool_diversity: 3 parameter_free: false --- > Methodology inspired by publicly taught earnings-preview frameworks; all text is an original paraphrase. ## Defaults | Parameter | Default Value | Rationale | |-----------|---------------|-----------| | surprise_window | 8 quarters | Standard surprise history window | | include_catalysts | true | Med prints move on catalysts as much as EPS | | guidance_sensitivity | true | Guidance is the swing factor for pharma | ## Preflight Run canonical pre-flight per `contracts/preflight.md`. Propagate X-Agentii-Trace per `contracts/x-agentii-trace-header.md`. ## Triggers - "Preview [biotech ticker]'s upcoming earnings." - "What should I expect at [ticker]'s next print?" - "How has [ticker] surprised historically?" - "Which catalysts land near [ticker]'s earnings date?" - "Build an earnings preview with consensus estimates." - "What's the guidance risk for [ticker] this quarter?" - "Summarize the last few quarters for [ticker]." - "Earnings + FDA calendar overlap for [ticker]." - "What are the swing factors for [ticker]'s print?" - "Historical reaction to [ticker]'s earnings surprises." - "What changed in my estimates, thesis, and positioning ahead of [ticker]'s print?" ## Production Grounding - Med prints have TWO drivers: financials (revenue/EPS/guidance) and catalysts (PDUFA/AdCom/readouts). The catalyst overlay is mandatory — a clean quarter can be undone by a CRL days earlier. - Every preview closes with model-vs-consensus deltas (our modeled numbers vs consensus from `search_earnings_calendar`, signed with the driver named) and the three what's-changed vectors (estimates / thesis / positioning) — estimates move first, thesis and positioning follow only when the facts justify them. - For pre-revenue biotechs, the print is mostly about cash runway + pipeline updates; consensus EPS is secondary. - Grounding frameworks: `references/knowledge-frameworks.md` (道/法 review knowledge). ## Data Source Priority 1. `search_earnings_calendar` — estimates, actuals, surprise history, next date. 2. `search_xbrl_facts` — revenue/EPS/margin trends. 3. `search_documents`/`read_source_*` — prior-quarter commentary and guidance. 4. Knowledge layer: `search_investment_cases` for historical print reactions. ## Methodology ### Retrieval Scope unstructured_document_search ### Retrieval Strategy 1. Resolve the earnings event: `search_earnings_calendar` for dates/estimates/surprises (consensus values). 2. Pull fundamentals trend: `search_xbrl_facts` key line items (our modeled numbers). 3. Compute model-vs-consensus deltas on each key line item; state assumptions; annotate coverage_gap where no consensus value exists. 4. Slot-refresh the three what's-changed vectors: estimates / thesis / positioning. 5. Catalyst overlay: nearest PDUFA/AdCom/readout vs print date. 6. Ground with historical cases (print reactions) via knowledge tools. ### Temporal Scope See frontmatter temporal_scope block. ### Tool Allowlist See frontmatter allowed_tools. ### Protocol 1. Event & estimates 2. Fundamental trend 3. Model-vs-consensus deltas 4. What's-changed vectors (estimates / thesis / positioning) 5. Catalyst overlay 6. Swing-factor synthesis ## Modes - **Standard** (default): estimates + surprises + guidance. - **Pre-revenue**: runway + pipeline + readout framing. - **Catalyst-overlap**: print framed around nearby FDA events. ## Tool Fallbacks | Failure | Fallback | |---------|----------| | search_earnings_calendar empty | Use filings (`search_documents`) for dates; annotate coverage_gap | | No catalyst data | Flag "catalyst overlay unavailable" | | Knowledge tools empty | Proceed with structured data only | ## Output File `{ticker}/{YYYY-MM-DD_HHMM}_earnings-preview-med_{affix}.md` ## Output Structure 1. **Executive Summary** — setup for the print in 2-3 sentences 2. **Consensus & Surprise History** — estimates table + surprise record 3. **Model vs Consensus** — our modeled numbers vs consensus (`search_earnings_calendar`) with signed deltas and named drivers 4. **What's Changed** — the three vectors: estimates / thesis / positioning (estimates move first, ratings last) 5. **Guidance & Swing Factors** — guidance risk analysis 6. **Catalyst Overlay** — FDA events near the print 7. **Historical Context** — cases with /v/ citations 8. **Coverage Gaps** — degraded flags ## Error Handling | Error | Fallback | |-------|----------| | Estimates missing | Present fundamentals trend only; flag | | Date uncertain | Use calendar's best estimate + annotate | ## Memory Load See `contracts/memory-load.md`. ## Snapshot See `contracts/snapshot-synthesis.md`. ## Final Summary (TUI) Include ### Key Citations block with 0-10 clickable /v/ URLs. ## References - `contracts/citation-and-memory.md` - `contracts/output-frontmatter-schema.md` - `contracts/memory-load.md` - `contracts/snapshot-synthesis.md` - `contracts/preflight.md` - `references/knowledge-frameworks.md`
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.