getaero-io/gtm-eng-skills / skills/deepline-gtm
deepline-gtm
UnverifiedGTM leads, enrichment, outreach, Plays, developer signals/Reo.dev. Providers: adyntel,affinity,ai_ark,akta,allegrow,amplemarket,apify,attention,attio,aviato,bettercontact,bigquery,bluesky,bounceban,brave,browserbase,builtwith,buyercaddy,cbinsights,clay,clickhouse,close,cloudflare,contactout,contextdev,crunchbase,crustdata-v3,customer_db,databricks,dataforseo,datagma,dealroom,deepline_ip_to_company,deepline_native,deeplineagent,deepseek,discolike,dnb,edges,emailbison,emailguard,enformion,enigma,exa,findymail,firecrawl,fireflies,firmable,forager,fullenrich,fundable,generic_http,getleads,gong,google_ads_audiences,google_workspace,govfiles,grain,hackernews,harvestapi,heyreach,hubspot,hunter,icypeas,instantly,intercom,ipqs,kaspr,kernel,kickbox,leadmagic,lemlist,limadata,linkedin_ads_audiences,luma,lusha,meerapi,meta_audiences,millionverifier,nooks,notion,openmart,opensosdata,openwebninja,outreach,parallel,particle,peopledatalabs,pitchbook,podscan,polymarket,postgres,predictleads,prospeo,quickenrich,rapidapi,….
references/monitor-operations.md
View on GitHub# Monitor operations
Lookup material for the [monitors recipe](../recipes/deepline-monitors.md):
deployment verification, monitor types and filters, commands, recovery paths and
the definition shape. Follow the recipe for setup, approval and observation.
## First proof of value: a minimally filtered scout
Do not make a customer wait days to discover whether the intended signal has
coverage. After approval, start the approved small scout batch with only the
target and requested event filters and the volume-based first window when
history is supported, then prove each stored definition after the write. This is the fastest feedback loop and guards against false success.
**Example outcome:** “Tell me when Stripe posts a Chief Financial Officer job,
then let a Play react.” Read the live tool contract before using this example.
### Optional: test the closest real signal
A paid current-signal probe is optional and needs approval of its live price.
Discover one with `deepline search "<tracked subject -> exact signal>"`;
prefer the monitor's provider. It must match both the event and its timing:
a contact's current profile does not prove a job change; a separate REST read
does not prove webhook delivery. If no faithful probe exists, say so. Inspect
identity, date and URL before calling a result a match. Empty coverage is
inconclusive, and current coverage does not prove future delivery.
### Verify every requested deployment
Read back every created/reactivated monitor. When `get` supplies `sample_payload`,
run the safe diagnostic in an internal/test workspace; obtain explicit approval
first in a customer workspace:
```bash
deepline monitors test <key> '<sample_payload from monitors get>' --json
```
Require `accepted: true`, `test_mode: validation_only`, `persisted_rows: 0`
and `dispatched_bound_plays: 0`. This checks event shape against the real binding,
not upstream emission. If no sample exists, report that no safe payload test is
available. Never fabricate a webhook body. A real event → DB → Play test needs
its own approved delivery path; `monitors test` has no dispatch mode.
```bash
# Learn the live job-opening payload fields, output stream, and event price.
deepline describe deepline_native.company_job_openings
# CFO openings at one company are sparse: start 180 days back.
UPDATES_SINCE=$(node -e 'console.log(new Date(Date.now() - 180 * 86400000).toISOString())')
MONITOR=$(node - "$UPDATES_SINCE" <<'JS'
console.log(JSON.stringify({
key: 'stripe-cfo-job-openings',
name: 'Stripe CFO job openings',
tool: 'deepline_native.company_radar',
payload: {
domain: 'stripe.com',
radar_type: 'company_job_openings',
job_titles: '"Chief Financial Officer"',
updates_since: process.argv[2],
},
}));
JS
)
# These are safe. They validate the exact definition and show cost/reuse.
deepline monitors check "$MONITOR"
deepline monitors deploy --dry-run "$MONITOR"
# After explicit approval of scope, shared-stream impact, and price:
deepline monitors deploy "$MONITOR"
deepline monitors get stripe-cfo-job-openings
# In an internal/test workspace, or after explicit customer approval:
deepline monitors test stripe-cfo-job-openings '<sample_payload from get>'
```
The deployment proof is the final `get` plus the safe test when available: the
definition must show `definition.payload.job_titles` as `"Chief Financial
Officer"`, the expected domain/radar type, and `status: active`; the safe test
must accept the provider-shaped sample without persistence or dispatch. If any
check fails, report a failed deployment. Do not wait for events to infer the
filter.
## Find monitor types and read their filters
Monitor types live on the `tools` surface, alongside every other capability.
Browse them, then read one type's exact filters + stream columns:
```bash
# Browse the monitor types you can deploy
deepline tools list --categories monitors
deepline search "company domains -> ongoing company event feed"
# Read one specific monitor variant's full contract
deepline describe deepline_native.company_job_openings
```
(`deepline monitors available [tool-id]` is a legacy alias for the same
discovery — prefer `tools`.)
The contract for a type gives you everything you need to deploy and to filter:
- **payload_schema** — the deploy-time filters you set in the monitor `payload`
(typed: required fields, allowed values).
- **stream columns** — the row fields a play filters on with `sqlListeners.where`
(the post-ingestion filter surface).
- **pricing** — live Deepline price, charge timing, and pricing basis.
- **a deploy example** — `deepline monitors deploy '<def>'`.
A monitor type is deployed with `deepline monitors deploy`, never run.
## Command set
Monitor commands use readable text by default; pass `--json` when parsing output.
Use `deepline describe <tool-id>` without `--json` for discovery.
| Command | Purpose |
| --------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- |
| `tools list --categories monitors` / `describe <tool-id>` | Discover live payload fields, streams and prices. |
| `monitors check '<definition>'` | Validate without deployment or spend. |
| `monitors preview '<definition>'` | Paid sample of recent matches, only where `check` returns `preview.available: true`; creates nothing. |
| `monitors deploy '<definition>'` | Apply a full desired definition; `--dry-run` previews validity, price and reuse. |
| `monitors list` | Page the registry with `--state` (`active` by default, or `all`), `--limit` and `--cursor`. |
| `monitors get <key>` | Read stored definition, current binding and `monitor_spec`. |
| `monitors test <key> '<payload>'` | Validate callback shape without rows, charges or Play dispatch. |
| `monitors validate <key>` | Validate provider configuration; does not prove delivery. |
| `monitors update <key> '<patch>'` | Apply a patch; may replace the resource. Validate the merged definition first. |
| `monitors delete <key>` | Delete upstream and local resources; supports `--dry-run`; requires `--yes` non-interactively. |
| `monitors reactivate <key>` | Resume a disabled monitor; inspect cost with `--dry-run`. |
| `monitors health [--fleet <id>]` | Inspect desired/provider state and delivery. Wait for `refresh.state: "completed"` before using whole-scope counts. |
| `monitors audit [--fleet <id>]` | Read upstream evidence; `--watch` waits for completion. Does not resume billing-paused ingestion. |
| `monitors repair [--fleet <id>]` | Converge observed state disagreements; audit and dry-run first. Does not bypass billing recovery. |
| `monitors fleets init` / `sync` / `get` | Define, converge or inspect a Fleet; `sync --dry-run` plans, `sync --wait` follows its durable operation. |
| `monitors fleets deactivate <id>` | Stop the owned Fleet; supports `--dry-run`, `--yes` and `--wait`. |
| `monitors fleets reactivate <id>` | Explicit recovery after credit suspension; inspect costs before `--yes --wait`. |
Prefix these commands with `deepline`. Use live `--help` for flags. Definition
commands accept positional JSON or `--file <path>` / `--file -`. When listing,
read `total`, `returned`, `has_more` and `next_cursor`; continue while `has_more`
is true. `--include-consumers` is limited to 20 rows. A top-up alone does not
recreate deleted Fleet resources.
## Preview samples
`monitors preview` returns up to three recent `updates` now and creates nothing.
Each returned match costs the per-match price; an empty preview is free and does
not prove zero matches. Keep each preview's key and `updates`: they are the
sample for the rest of setup, because a preview stores no rows. Deploy the
approved definition, without `--limit`, at the destination step.
To send a preview through the action, run the Play once per update that passes
the listener's `where` conditions, which you apply yourself:
```json
{
"tool": "<monitor tool>",
"stream": "<stream>",
"operation": "INSERT",
"before": null,
"after": {
"domain": "<update.data.domain>",
"discovered_at": "<update.discovered_at>",
"<each update.content field that is a column of the stream>": "..."
},
"changedAt": "<update.discovered_at>",
"metadata": {
"outboxId": "sample:<preview key>:<update index>",
"listenerId": "sample",
"table": "<schema.table>"
}
}
```
These events lack `event_id`, `record_id`, `radar_id` and every `_dl_*` column. If
the Play reads one of them, say the sample can't exercise that step; don't invent
a value.
## Billing pause: current safe operator path
`status: active` is control-plane intent, not delivery proof; start with
`health` when a feed is quiet. `audit` records provider evidence and `repair`
converges existing intent. Neither resumes a zero-credit pause.
There is no org-wide direct-monitor resume. After credits or a cap are fixed:
1. Read `deepline monitors health` until it no longer reports refreshing.
2. Reactivate a credit-suspended Fleet through its retained owner:
```bash
deepline monitors fleets reactivate <fleet-id> --yes --wait
```
3. For a fully disabled direct monitor, read its cost before recovery:
```bash
deepline monitors reactivate <key> --dry-run
deepline monitors reactivate <key>
```
Do not fan out direct recovery while suspension is still stopping resources: it
can race delete/create. Report that gap rather than inventing bulk recovery.
## Recover from errors by code
Read the returned state before retrying. For validation errors, correct every
reported field against the live contract and rerun `check`. For insufficient
credits, report the required credits, balance, and shortfall, then stop. Retry a
transient read or check once; do not blindly repeat a create, settlement, or
cleanup failure.
## Monitor definition shape
A definition is a single JSON object:
```json
{
"key": "company-job-openings",
"tool": "deepline_native.company_radar",
"name": "Company job openings",
"payload": {
"domain": "stripe.com",
"radar_type": "company_job_openings"
},
"controls": {}
}
```
- `key` — public monitor instance id (you reference it in `get`/`update`/`delete`).
- `tool` — a live Deepline-native tool id. Get the valid ids and each
`payload_schema` from `deepline tools list --categories monitors` /
`deepline describe <tool-id>`.
- `payload` — tool-specific; must match that tool's `payload_schema`.
- `name` — optional human label. `controls` — optional Deepline lifecycle metadata.
The same object is what `defineMonitor({ ... })` returns (typed) and what
`client.monitors.check`/`deploy` accept — the CLI JSON and the SDK definition are
one shape. See "Monitors as code (SDK)".
### Priority radar exception
Use `controls.execution_type: "priority"` only for an urgent preview or
calibration, such as the setup sample radar, and never for the targets turned on
after it. Deepline adds the provider-facing marker; do not put
`custom_fields` in `payload`. Regular and bulk creation stay normal. To keep a
stopped radar but return it to regular execution, patch
`{"controls":{"execution_type":null}}`.
```json
{
"key": "stripe-job-openings-preview",
"tool": "deepline_native.company_radar",
"payload": {
"domain": "stripe.com",
"radar_type": "company_job_openings"
},
"controls": { "execution_type": "priority" }
}
```
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |