- Triggers
- which accounts are at riskshould we bid on this RFPpipeline coverageforecast accuracyGTM efficiencychurn risk
- Best for
- For sales execution, customer success, and contracts. Pick the dominant intent (health score, RFP response, pipeline math, proposal draft).
- Guardrail
- Not for per-deal pricing / discount / forecast — those go to commercial-orchestrator. Not for strategic CRO judgment — escalate to c-level-advisor-router.
Clara / Live dashboard
Interactive demoMy personal AI assistant
Clara coordinates the work.
I own the decisions.
Explore the 14 specialized agents behind my finance, data, research, reporting, and quality-control workflows. Open any card to see its role, triggers, ideal use, and guardrails.
- Agents
- 14
- Capabilities
- 6
- Active
- 7
- Review
- Human
01Your brief → 02Clara routes → 03Jill reviews
Agent roster
14 agents · 6 capabilities
7 active · 6 thinking · every final output reviewed by Jill
Capability
Status
Showing 14 of 14 agents
- Triggers
- I need a CEO perspectiveCFO reviewCISO reviewterm sheetadvisor for Xfounder-mode
- Best for
- For founder-mode questions, term sheets, M&A, layoffs, fundraising, multi-domain trade-offs, irreversible decisions. Reads ~/.claude/company-context.md first if present.
- Guardrail
- Never auto-decides. Every recommendation names the human owner. Tactical single-domain work belongs to a specialist.
- Triggers
- review this dealshould we discountpricing modelpartner economicsRFP responsebookings forecast
- Best for
- For per-deal pricing, discount review, RFP/RFI response, partnership tier, channel ROI, bookings forecast. Voice: margin at full discount AND next-quarter pipeline at the same terms.
- Guardrail
- Not for sales execution (business-growth-router). Pricing outputs are model + range, never a specific number — the human picks.
- Triggers
- which compliance frameworks applyaudit calendarSOC 2 + ISO 27001 overlapevidence reusemock auditISO 42001 readiness
- Best for
- For multi-framework programs (2+ frameworks). Runs 4 decisions: framework selection, cross-framework mapping, mock audit, unified evidence checklist.
- Guardrail
- Not a per-framework deep-dive — routes per-framework work to ra-qm-team, gdpr-dsgvo-expert, etc. Cross-walks are advisory; novel ones need counsel.
- Triggers
- help me scaffoldreview my codeset up CI/CDdesign a schemafine-tuneWCAG audit
- Best for
- Front door for any software engineering question without an obvious specialist. Picks one specialist and loads only that SKILL.md. Never bulk-loads.
- Guardrail
- When the task needs depth (RAG, MCP, SLO design, chaos engineering, kubernetes operator), delegate to engineering-powerful-router.
- Triggers
- design an agentbuild RAGbuild an MCP serverdesign an SLOchaos experimentfeature-flag rollout
- Best for
- For architecture / automation / reliability / platform-ops depth. Pairs with engineering-core-router (e.g., senior-backend + api-design-reviewer).
- Guardrail
- Specialist-level — does not run meta-harness loops. For pure coding roles without architecture depth, prefer engineering-core-router.
- Triggers
- audit this Excel fileprofile this workbookscan project for Excel anti-patternsisolated Excel deep-dive
- Best for
- Dispatch this agent only when the parent wants a background audit that should NOT pollute the main conversation. Multi-sheet workbooks, hundreds of formulas, project-wide scans.
- Guardrail
- For normal Excel questions, the excel skill handles inline. Domain-specific — nothing outside Excel.
- Triggers
- analyze these financialshow healthy are my SaaS metricsDCF valuationbudget vs actualARR growthCAC payback
- Best for
- Pick the lens: financial-analyst (statement-level) vs saas-metrics-coach (ARR/MRR/churn/CAC/LTV/NRR). Outputs are analysis support, not investment advice.
- Guardrail
- Not for strategic CFO judgment (c-level-advisor → CFO). Not for per-deal economics (commercial-orchestrator). Not for R&D program finance (research-ops → research-finance).
- Triggers
- create a landing pagebuild a landing pagemake a landing page for XI need a web page for Ypromotional pageproduct page
- Best for
- When you want a premium single-file HTML landing page with animation. Runs 4-Q grill-me intake (product/audience/brand/tone) → single self-contained .html file.
- Guardrail
- Not for Next.js/TSX (use product-team/landing-page-generator). Output is one .html file — all CSS inline, only Google Fonts + GSAP via CDN.
- Triggers
- find a loopaudit this loopdesign a recurring agentreview my automationthis loop runs foreverturn this into a loop
- Best for
- For any recurring agent / automation cadence problem. Treats a loop as a feedback system with terminal states, not as permission for endless autonomy.
- Guardrail
- Never invents Loop Library titles/numbers/URLs. Labels adaptations as such. If the live catalog is unavailable, says so plainly — does not substitute repo content.
- Triggers
- convert this markdown to HTMLmake this an HTML fileturn this into an interactive documentrender this report as HTMLPR writeup as HTMLslides from this markdown
- Best for
- For long-form documents, code reviews, or slide decks from markdown. Voice: "What decision does this HTML drive — is the reader skimming, deciding, or presenting?"
- Guardrail
- Refuses input < 100 lines (markdown still wins). Refuses if design-system not onboarded. Not a landing-page generator. Single-file HTML only.
- Triggers
- help me prioritizeplan a product experimentwe ship features nobody usesrun the discovery loopis our OST soundwrite a PRD
- Best for
- Routes across 16 product lanes (RICE, OKRs, UX research, competitive teardown, analytics, experiments, discovery, roadmaps, spec-to-repo) AND runs the weekly discovery cadence with verifiable gates.
- Guardrail
- Not the delivery loop — that is project management. The OST linter is binding: no roadmap item may cite a failing tree.
- Triggers
- brain dumptriage my inboxweekly reviewplan a deep work blockprep for a meetingextract action items
- Best for
- Pick the dominant intent — capture, email, fable-goal, andreessen, meetings, weekly-review, deep-work, reflect, roast, or handoff. Finish one, then route again.
- Guardrail
- Specialist gates are binding — weekly-review refuses without GET-CURRENT; deep-work refuses > 4h deep demand; meetings enforces cost gate. The agent refuses overrides.
- Triggers
- design this clinical studywhat sample sizeR&D budgetburn ratecapitalize or expenseTAM SAM SOM
- Best for
- Routes across 4 R&D lanes (clinical-research, research-finance, market-research, product-research). Voice: "What decision does this research drive, and what is your confidence — show me the method and the assumptions before the number."
- Guardrail
- Clinical outputs are estimates + named clinical owner (never fact). Market size shows method + both-ways triangulation (never a single unsourced number). Product insights require recurrence across independent participants.
No agents match the current filters.
Interactive portfolio view. Statuses and tasks are representative examples of how Clara routes work; Clara does not operate without Jill's review and final approval.