Mode	Signal	Framework
STRATEGY	Market entry, partnerships, resource allocation, "should we"	FRAME -> ANALYZE -> DECIDE
TECHNOLOGY	Build vs buy, vendor, SaaS, tech stack	TCO analysis -> Vendor scoring -> Decision
GROWTH	Audience, SEO, brand, community, channel	Segment -> Channel-score -> Plan
COMPETITIVE	Competitor, market landscape, differentiation	Map landscape -> Position -> Track
EVALUATION	Feasibility, ROI, go/no-go, effort estimate	Feasibility score -> ROI model -> Verdict
SUPPORT	Ticket triage, response, KB article, escalation	TRIAGE/RESPOND/KB/ESCALATE/RESEARCH
FINANCE	Journal entry, reconciliation, variance, SOX	CLASSIFY -> GATHER -> PRODUCE -> VERIFY
HR	Recruiting, performance review, compensation, hiring	DEFINE -> PIPELINE/EVALUATE -> DOCUMENT
LEGAL	Contract review, compliance, NDA, DSGVO, GoBD	INTAKE -> ANALYZE -> FLAG -> OUTPUT
OPERATIONS	Runbook, process docs, risk, vendor, change mgmt	SCOPE -> AUTHOR/ASSESS -> VERIFY
SALES	Call prep, pipeline, outreach, forecast	RESEARCH -> PREPARE/ANALYZE -> OUTPUT
PRODUCTIVITY	Task management, daily plan, meeting, standup	CAPTURE -> DECOMPOSE -> PRIORITIZE
PRODUCT	Feature spec, PRD, roadmap, user research, metrics	UNDERSTAND -> GATHER -> GENERATE

If the request spans modes, pick the primary. Note the secondary.

Shared Workflow

Classify the request into exactly one mode from the table.
Load the mode's reference file (see Deep References). Each reference contains full framework, sub-modes, phases, and deeper reference pointers.
Follow the reference's instructions. Do not improvise a framework when one exists.
Every mode's top-level reference also specifies an LLM failure-modes guardrail file. Load it alongside the main reference.

Executive Modes (STRATEGY, TECHNOLOGY, GROWTH, COMPETITIVE, EVALUATION)

Load references/csuite.md for the full C-suite framework. All five modes share the same 3-phase pattern:

Phase 1: FRAME -- Convert the question into a structured decision. Name the actual decision (users present symptoms; the real decision is broader). Identify irreversibility. Set time horizon. List 2-4 options. State what makes it hard.

Phase 2: ANALYZE -- Per option: upside (best realistic + expected), downside (worst realistic + recovery + irreversible losses), requirements, opportunity cost. Separate facts from assumptions. Quantify.

Phase 3: DECIDE -- Score options via weighted criteria. Present verdict with driving factors, risk mitigations, and reversibility assessment. Load mode-specific deep references for scoring matrices.

Support Mode

Load references/customer-support.md. Five sub-modes: TRIAGE (classify ticket, assign priority P1-P4, route), RESPOND (draft calibrated reply with tone matching emotional state), KB (convert resolution into publish-ready article), ESCALATE (build structured brief with impact assessment), RESEARCH (investigate history with confidence scoring).

Key guardrail: never fabricate ticket history, resolution steps, or customer quotes.

Finance Mode

Load references/finance.md. Six sub-modes: JOURNAL ENTRY, RECONCILIATION, VARIANCE, STATEMENTS, AUDIT/SOX, CLOSE. All output is working material for qualified professionals -- not financial advice.

Key guardrails: never fabricate account codes, balances, or transaction details. Source all numbers from user-provided data. Always state GAAP/IFRS basis.

HR Mode

Load references/hr.md. Nine sub-modes: RECRUITING, PERFORMANCE, COMPENSATION, OFFER, INTERVIEW, ONBOARDING, ORG-PLANNING, PEOPLE-ANALYTICS, POLICY.

Key guardrails: source compensation data from user-provided or public databases. Focus on skills, behaviors, outcomes -- not demographics. Include legal review disclaimer on binding language. Ask for jurisdiction before compliance advice. Minimize PII retention.

Legal Mode

Load references/legal.md. Six sub-modes: CONTRACT (clause-by-clause GREEN/YELLOW/RED analysis), COMPLIANCE (regulation checklist with German compliance for DSGVO/GoBD/TDDDG), NDA (triage classification), RISK (severity x likelihood matrix), WRITING (structured legal documents), VENDOR (agreement inventory and gap analysis).

Key guardrail: analysis support only, not legal advice. All binding language needs qualified counsel review. Load references/legal/german-business-compliance.md for German-specific regulation.

Operations Mode

Load references/operations.md. Nine sub-modes: RUNBOOK, RISK, VENDOR, PROCESS, CHANGE, CAPACITY, COMPLIANCE, STATUS, OPTIMIZE.

Key patterns: RUNBOOK uses SCOPE -> AUTHOR -> VERIFY with prerequisite checklists. RISK produces severity x likelihood matrices. CHANGE follows RFC -> REVIEW -> IMPLEMENT -> VERIFY. All operations docs require rollback procedures.

Sales Mode

Load references/sales.md. Seven sub-modes: CALL-PREP, PIPELINE, OUTREACH, COMPETITIVE, FORECAST, CALL-SUMMARY, RESEARCH.

Key patterns: CALL-PREP researches prospect then builds agenda with questions. PIPELINE applies health scoring and flags stale deals. OUTREACH personalizes from research (never generic templates). FORECAST uses weighted probability with scenario analysis.

Productivity Mode

Load references/productivity.md. Six sub-modes: TASK (decompose and prioritize), PLAN (daily/weekly time blocks), MEETING (optimize or eliminate), STATUS (structured updates), REVIEW (weekly retrospective), GOAL (OKRs).

Key patterns: vertical slicing for task decomposition (shippable slices, not horizontal layers). 1/2/4-hour time buckets. Tasks over 4 hours need decomposition.

Product Mode

Load references/product-management.md. Eight sub-modes: SPEC (PRD with acceptance criteria), ROADMAP (Now/Next/Later prioritization), STAKEHOLDER (structured updates), RESEARCH (synthesis with thematic analysis), COMPETITIVE (battle cards), METRICS (funnel/cohort/retention analysis), SPRINT (backlog grooming with capacity), BRAINSTORM (Socratic exploration).

Error Handling

Error	Response
Too many options (5+)	Eliminate obviously inferior first. Get to 2-4.
Not enough information	Identify 2-3 critical unknowns. Recommend research sprint.
Analysis paralysis	Reversibility test. If reversible, recommend best option with checkpoint.
Emotional attachment	Name the pattern. Ask: stress-test, or genuinely evaluate?

Deep References

All references contain >100 lines of domain-specific frameworks, templates, and guardrails. Load on demand per mode.

Mode	Primary	Sub-references
STRATEGY	`references/csuite.md`	`references/strategic-frameworks.md`, `references/decision-matrices.md`
TECHNOLOGY	`references/csuite.md`	`references/tco-framework.md`, `references/vendor-evaluation.md`
GROWTH	`references/csuite.md`	`references/audience-segmentation.md`, `references/channel-evaluation.md`
COMPETITIVE	`references/csuite.md`	`references/competitive-mapping.md`, `references/market-positioning.md`
EVALUATION	`references/csuite.md`	`references/feasibility-scoring.md`, `references/roi-frameworks.md`
SUPPORT	`references/customer-support.md`	`references/customer-support/*.md`
FINANCE	`references/finance.md`	`references/finance/*.md`
HR	`references/hr.md`	`references/hr/*.md`
LEGAL	`references/legal.md`	`references/legal/*.md`
OPERATIONS	`references/operations.md`	`references/operations/*.md`
SALES	`references/sales.md`	`references/sales/*.md`
PRODUCTIVITY	`references/productivity.md`	`references/productivity/*.md`
PRODUCT	`references/product-management.md`	`references/product-management/*.md`
Cross-cutting	`references/content-funnel.md`, `references/trend-analysis.md`, `references/risk-assessment.md`, `references/estimation-techniques.md`, `references/migration-planning.md`	Shared frameworks

business-ops

Business Operations

Mode Detection