Enterprise Signal Listener — Live Signal Capture
Extract everything needed to build an enterprise AI agent from natural conversation — without
interrupting the person talking. Capture while they speak. Ask nothing unless a critical signal
is missing.
Core Behavior
Do not ask the person to fill out a form. Do not pause the conversation to confirm understanding.
Listen, extract, and produce a completed intake specification. Ask at most two clarifying
questions, and only when a critical signal cannot be inferred.
What to Listen For
Extract these signals from natural speech. They will rarely be stated explicitly — infer them
from context:
| Signal Type |
Listen for phrases like... |
| Platform |
"in ServiceNow", "our Salesforce", "Workday tickets", "SAP transactions", "Snowflake warehouse", "Teams channel", "Azure pipeline", "our AWS account", "Okta provisioning", "Splunk alerts", "CrowdStrike detections", "Epic orders", "Veeva submissions", "Bloomberg feed" |
| Industry |
"patients", "clinical trials", "adverse events", "trades", "positions", "claims", "underwriting", "SKUs", "warfighters", "compliance audit", "drug approval", "loan origination", "grid operators", "grid assets", "procurement", "purchase orders" |
| Pain trigger |
"manually", "takes too long", "we keep missing", "no one knows when", "we get paged at 3am", "the team wastes hours on", "audit keeps failing", "we can't see", "there's no visibility into", "someone has to", "it falls through the cracks" |
| Actor |
Who does this today? (a person, a named team, a queue, a shift) |
| Frequency |
How often does this happen? ("every ticket", "nightly", "on-demand", "real-time", "every quarter", "whenever a new hire joins") |
| Outcome wanted |
"automatically", "instantly notify", "alert us when", "summarize for us", "route it to", "predict when", "remediate it", "generate the report", "approve without a human", "escalate if" |
Rapid Intake Form
Populate this from conversation signals. Mark fields [inferred] if derived from context,
[missing] if not detectable:
intake:
spoken_problem: "" # verbatim quote or close paraphrase of what they said
platform_hint: "" # any system(s) they mentioned — be specific (e.g. "ServiceNow ITSM")
industry_hint: "" # domain or vertical (e.g. "healthcare — hospital network")
actor: "" # who does this today (person, team, queue)
trigger: "" # what kicks the process off (event, schedule, threshold, request)
desired_outcome: "" # what "fixed" or "done" looks like in their words
pain_intensity: "" # low | medium | high | critical (infer from language: "3am", "audit fails" = critical)
urgency: "" # demo_now | this_week | this_sprint | next_quarter
compliance_flags: [] # any regulatory or compliance terms heard (HIPAA, FedRAMP, SOX, PCI, GxP, etc.)
secondary_systems: [] # any additional platforms or integrations mentioned
Clarifying Questions
Ask at most 2 questions. Only ask if a critical signal is missing after careful listening.
Stop as soon as you have enough to hand off to problem decoding.
Priority order — ask the first unanswered question from this list:
- "Which system does this live in today?" — ask only if no platform was detectable from the conversation
- "When this goes wrong, who feels it first?" — surfaces the actor and trigger simultaneously
- "What would 'fixed' look like in 30 seconds?" — surfaces the desired outcome
Never ask about:
- Timelines, budgets, or project schedules
- Technical architecture or infrastructure preferences
- Whether they've tried other solutions
- Org structure or reporting lines
Build from what you have. Ambiguity is resolved in the Decode phase, not here.
Signal Quality Assessment
Before handing off, rate the intake:
| Field |
Status |
Action if missing |
spoken_problem |
Required |
Paraphrase what was described |
platform_hint |
Required |
Ask clarifying question #1 |
industry_hint |
Preferred |
Infer from domain vocabulary; flag as [inferred] |
actor |
Preferred |
Ask clarifying question #2 |
trigger |
Preferred |
Infer from frequency and pain language |
desired_outcome |
Required |
Ask clarifying question #3 |
pain_intensity |
Auto-inferred |
Always derivable from language |
urgency |
Preferred |
Default to this_week if not stated |
If spoken_problem, platform_hint, and desired_outcome are all populated — even with
inferred values — the intake is complete and ready for decoding.
Output
Produce the completed intake YAML and a one-sentence summary:
SIGNAL CAPTURE COMPLETE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Summary: [One sentence describing the problem heard, the platform, and the desired outcome]
Intake:
[populated intake YAML]
Signal quality: [Complete | Partial — missing: field1, field2]
Ready for: enterprise-problem-decoder
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Hand off the completed intake YAML to enterprise-problem-decoder.
1---2name: enterprise-signal-listener3description: Use this skill when someone — a client, stakeholder, investor, team member, or executive — is describing a business problem, workflow pain, or opportunity in natural conversation and you need to extract structured signal without interrupting them. Activate the moment someone says "we need something that can..." or "our team keeps having to manually..." or "what if there were a way to..." Produces a completed intake specification ready for problem decoding. This is Phase 0 (Listen) of the Listen → Decode → Build → Ship methodology.4license: Apache-2.05---67# Enterprise Signal Listener — Live Signal Capture89Extract everything needed to build an enterprise AI agent from natural conversation — without10interrupting the person talking. Capture while they speak. Ask nothing unless a critical signal11is missing.1213---1415## Core Behavior1617Do not ask the person to fill out a form. Do not pause the conversation to confirm understanding.18Listen, extract, and produce a completed intake specification. Ask at most two clarifying19questions, and only when a critical signal cannot be inferred.2021---2223## What to Listen For2425Extract these signals from natural speech. They will rarely be stated explicitly — infer them26from context:2728| Signal Type | Listen for phrases like... |29|-------------|---------------------------|30| **Platform** | "in ServiceNow", "our Salesforce", "Workday tickets", "SAP transactions", "Snowflake warehouse", "Teams channel", "Azure pipeline", "our AWS account", "Okta provisioning", "Splunk alerts", "CrowdStrike detections", "Epic orders", "Veeva submissions", "Bloomberg feed" |31| **Industry** | "patients", "clinical trials", "adverse events", "trades", "positions", "claims", "underwriting", "SKUs", "warfighters", "compliance audit", "drug approval", "loan origination", "grid operators", "grid assets", "procurement", "purchase orders" |32| **Pain trigger** | "manually", "takes too long", "we keep missing", "no one knows when", "we get paged at 3am", "the team wastes hours on", "audit keeps failing", "we can't see", "there's no visibility into", "someone has to", "it falls through the cracks" |33| **Actor** | Who does this today? (a person, a named team, a queue, a shift) |34| **Frequency** | How often does this happen? ("every ticket", "nightly", "on-demand", "real-time", "every quarter", "whenever a new hire joins") |35| **Outcome wanted** | "automatically", "instantly notify", "alert us when", "summarize for us", "route it to", "predict when", "remediate it", "generate the report", "approve without a human", "escalate if" |3637---3839## Rapid Intake Form4041Populate this from conversation signals. Mark fields `[inferred]` if derived from context,42`[missing]` if not detectable:4344```yaml45intake:46 spoken_problem: "" # verbatim quote or close paraphrase of what they said47 platform_hint: "" # any system(s) they mentioned — be specific (e.g. "ServiceNow ITSM")48 industry_hint: "" # domain or vertical (e.g. "healthcare — hospital network")49 actor: "" # who does this today (person, team, queue)50 trigger: "" # what kicks the process off (event, schedule, threshold, request)51 desired_outcome: "" # what "fixed" or "done" looks like in their words52 pain_intensity: "" # low | medium | high | critical (infer from language: "3am", "audit fails" = critical)53 urgency: "" # demo_now | this_week | this_sprint | next_quarter54 compliance_flags: [] # any regulatory or compliance terms heard (HIPAA, FedRAMP, SOX, PCI, GxP, etc.)55 secondary_systems: [] # any additional platforms or integrations mentioned56```5758---5960## Clarifying Questions6162Ask **at most 2** questions. Only ask if a critical signal is missing after careful listening.63Stop as soon as you have enough to hand off to problem decoding.6465**Priority order — ask the first unanswered question from this list:**66671. "Which system does this live in today?" — ask only if no platform was detectable from the conversation682. "When this goes wrong, who feels it first?" — surfaces the actor and trigger simultaneously693. "What would 'fixed' look like in 30 seconds?" — surfaces the desired outcome7071**Never ask about:**72- Timelines, budgets, or project schedules73- Technical architecture or infrastructure preferences74- Whether they've tried other solutions75- Org structure or reporting lines7677Build from what you have. Ambiguity is resolved in the Decode phase, not here.7879---8081## Signal Quality Assessment8283Before handing off, rate the intake:8485| Field | Status | Action if missing |86|-------|--------|------------------|87| `spoken_problem` | Required | Paraphrase what was described |88| `platform_hint` | Required | Ask clarifying question #1 |89| `industry_hint` | Preferred | Infer from domain vocabulary; flag as `[inferred]` |90| `actor` | Preferred | Ask clarifying question #2 |91| `trigger` | Preferred | Infer from frequency and pain language |92| `desired_outcome` | Required | Ask clarifying question #3 |93| `pain_intensity` | Auto-inferred | Always derivable from language |94| `urgency` | Preferred | Default to `this_week` if not stated |9596If `spoken_problem`, `platform_hint`, and `desired_outcome` are all populated — even with97inferred values — the intake is complete and ready for decoding.9899---100101## Output102103Produce the completed intake YAML and a one-sentence summary:104105```106SIGNAL CAPTURE COMPLETE107━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━108109Summary: [One sentence describing the problem heard, the platform, and the desired outcome]110111Intake:112[populated intake YAML]113114Signal quality: [Complete | Partial — missing: field1, field2]115Ready for: enterprise-problem-decoder116━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━117```118119Hand off the completed intake YAML to `enterprise-problem-decoder`.
aviskaar