Your work. Your standards.

Run your next task.

Give an agent the context, working instructions, and quality bar your team actually needs. Create a task, run it, and inspect the work.

Opening your task workspace…

Start with real work

Bring the context. Set your standards. Each template creates a private task with a working protocol, a review rubric, and checks you can run again.

Customer support

Resolve a customer escalation

Turn a difficult ticket into an evidence-backed reply, an internal handoff, and a clear next step.

You bring: ticket, account context, and support policy

Deliverables & review criteria for Resolve a customer escalation

What the agent produces

  • Customer reply
  • Internal handoff
  • Evidence and uncertainties
  • Next actions

How the result is reviewed

Every material claim has evidence
The evidence section maps the reply's diagnosis, impact, and workaround claims to source line citations. Unverified causes and workarounds are explicitly qualified.
No unauthorized commitments
The reply follows supplied tone and policy rules, makes no unsupported refund or resolution promise, and does not expose internal notes or unnecessary personal information.
The customer knows what to do next
The reply acknowledges the specific business impact, gives a supported next step or a precise information request, and separates confirmed workarounds from untested suggestions.
The next teammate can act
The handoff contains the observed symptom, relevant identifiers, attempted steps, missing evidence, and proposed escalation destination without inventing ownership or urgency policy.
Engineering & operations

Write an incident postmortem

Reconstruct an incident from scattered notes and produce a blameless review with testable follow-up actions.

You bring: incident timeline, logs, and team review rules

Deliverables & review criteria for Write an incident postmortem

What the agent produces

  • Executive summary
  • Impact and timeline
  • Cause analysis
  • Corrective actions
  • Open questions

How the result is reviewed

A traceable, consistent timeline
Each material timeline event cites source lines. Timezones and interval calculations are explicit; missing or conflicting timestamps are flagged rather than silently reconciled.
Impact stays within the evidence
The review distinguishes observed affected users or requests from unknown totals and does not claim no data loss, full recovery, or a severity level without supporting evidence or supplied policy.
Hypotheses are not presented as root cause
Confirmed facts, suspected triggers, contributing conditions, and unanswered causal questions are separate. Temporal correlation alone is not treated as proof.
Follow-up actions have acceptance checks
Every action addresses an evidenced gap and includes a concrete verification method plus a sourced owner and date or explicit Unassigned and Not agreed values.
Product & research

Turn customer feedback into product decisions

Find patterns across interviews and tickets, preserve disagreement, and turn evidence into a prioritized decision brief.

You bring: customer feedback and product context

Deliverables & review criteria for Turn customer feedback into product decisions

What the agent produces

  • Decision brief
  • Themes and evidence
  • Disagreements and limitations
  • Prioritized opportunities
  • Next research steps

How the result is reviewed

Themes preserve the source meaning
Every theme has source line citations and a clear customer problem. Quotes are exact, and suggestions are not misrepresented as validated needs or requirements.
Counts use explicit units
Counts distinguish feedback items, distinct accounts, and unknown identities. Duplicate account evidence is not double-counted as independent support; sample counts are not generalized to the market.
Contradictions and gaps remain visible
The brief preserves minority or conflicting feedback, identifies selection and missing-context limits, and does not manufacture consensus.
Priorities explain their evidence and uncertainty
Each proposed priority follows the supplied decision rules or an explicitly provisional rationale, cites supporting evidence, and names an unanswered question that could change the ranking.
Procurement & operations

Compare vendors against your requirements

Turn proposals and internal requirements into a defensible shortlist with explicit tradeoffs and diligence questions.

You bring: requirements, vendor proposals, and decision rules

Deliverables & review criteria for Compare vendors against your requirements

What the agent produces

  • Recommendation and conditions
  • Requirements matrix
  • Cost comparison
  • Risks and tradeoffs
  • Diligence questions

How the result is reviewed

Every requirement has a sourced status
The matrix covers supplied must-haves and preferences for every vendor, cites source lines, and uses Meets, Does not meet, or Unresolved without treating silence as compliance.
Comparable costs have visible arithmetic
Costs use an explicit currency, seat count, billing period, and formula. Unquoted add-ons, taxes, implementation, and renewal terms remain unknown rather than being silently priced at zero.
Claims are distinguished from commitments
The analysis distinguishes sales claims, supplied contractual terms, and verified evidence, and preserves contradictions or missing diligence on mandatory requirements.
The shortlist follows the decision rules
The recommendation follows supplied constraints and weights, makes unresolved blockers explicit, and avoids a definitive winner when essential evidence is missing. Each diligence question targets a decision-changing gap.
Product & engineering

Review a launch before the go/no-go meeting

Check release evidence against your launch gates and produce a decision memo with blockers, rollback gaps, and owners.

You bring: launch plan, release evidence, and go/no-go criteria

Deliverables & review criteria for Review a launch before the go/no-go meeting

What the agent produces

  • Go/no-go recommendation
  • Gate review
  • Blockers and accepted risks
  • Rollout and rollback checks
  • Decision agenda

How the result is reviewed

Every required gate is reviewed
Each supplied launch gate has a Ready, Blocked, or Unverified status backed by source line citations. Planned work or unchecked assertions do not count as completed verification.
Risk acceptance has an actual owner
Blockers, unknowns, and accepted risks are separate. Accepted risks cite documented approval and authority; absent sign-offs, owners, and dates are explicitly missing.
Rollout and rollback gaps are actionable
The review checks rollout scope, health thresholds, stop conditions, monitoring ownership, and rollback evidence, and names the missing check needed for each unresolved area.
The recommendation matches the gates
The proposed go/no-go decision follows the supplied policy, does not override unmet mandatory gates, and includes a concise meeting agenda with decision owners or explicit owner gaps.

Loading tasks and run presets…

Follow the evidence behind your run

Find the tasks, data, and results that move your work forward.

Workflow illustrations — sample scenes, not live activity or measured results.

Trajectories

Inspect an agent’s tool calls, responses, and outcomes, step by step.
Workflow illustration
Explore trajectories

Tasks

Explore tasks with clear instructions, environments, and verifiable outcomes.
Workflow illustration
Explore tasks

Trials

Track your recent trials and their status.
Workflow illustration
View your runs