Signal Detection Pipeline

Signal Detection Pipeline

A thin orchestration over signal sub-skills. Run only the sources relevant to the client's ICP — each is independent and parallelizable. Companies appearing across multiple signals are the highest-confidence leads. (Pairs with signal-scanner for TAM-scoped, scheduled, storage-backed scanning.)

When to use

• "Find companies that might need [our product]." / "Run signal detection for [problem area]."
• "Find buying signals in [industry/topic]."

Sub-skills orchestrated (run relevant ones in parallel)

1. Job postings — job-posting-intent (Apify jobs / keyless serp).
2. Funding — funding-signal-monitor (serp + HN + optional Apify).
3. Event attendance — luma-event-attendees (web-automation / Apify Luma).
4. Reddit pain — keyless site:reddit.com web search (Extract Content); depth via an Apify Reddit actor (optional).
5. LinkedIn content — the LinkedIn engager skills (competitor-post-engagers / pain-language-engagers) via Phantombuster, or web-automation.

Then:

6. Dedup — contact-cache / a datatable dedup on company; tag the signal stack per company. Multi-signal = strongest leads.
7. Qualify + context — lead-qualification + the agent writes the outreach angle per qualified company.
8. Output — the ranked combined list + Agent Teams channel attachment.

How to run

Invoke each relevant sub-skill's scripts by path and merge their outputs in-agent, e.g.:

# 1. job-posting intent
python3 ${SKILL_DIR}/../job-posting-intent/scripts/jobs.py --query "Data Engineer" \
  --output ${WORKSPACE}/jobs.json
python3 ${SKILL_DIR}/../job-posting-intent/scripts/aggregate_intent.py \
  --input ${WORKSPACE}/jobs.json --intent-titles "data engineer" --output ${WORKSPACE}/job_co.json

# 2. funding
python3 ${SKILL_DIR}/../funding-signal-monitor/scripts/funding_search.py \
  --industries "devtools" --output ${WORKSPACE}/funding.json

# 4. reddit pain (keyless)
python3 ${SKILL_DIR}/../funding-signal-monitor/scripts/funding_search.py --help  # (pattern ref)

The agent merges per-company signals, dedups, and tags the signal stack. Each sub-skill owns its proxy/rate-limit/cost handling — degrade per its notes when an optional key is missing.

Outputs

A ranked, deduplicated lead list with per-company signal stack and outreach context (company, signals, score, angle) + Agent Teams channel attachment.

Credentials / env

• Required: none at the playbook level — gates live in the sub-skills it runs.
• Optional (per sub-skill): APIFY_API_TOKEN (jobs/Luma/Reddit depth), PHANTOMBUSTER_API_KEY + cookie (LinkedIn content), SUPABASE_URL + SUPABASE_SERVICE_ROLE_KEY (dedup/history), ANTHROPIC_API_KEY if not platform-provided.

Notes & edge cases

• Run only the sources relevant to the client's ICP; each is independent and parallelizable.
• Companies across multiple signals are the highest-confidence leads — rank them up.
• Each sub-skill owns its proxy/rate-limit/cost handling; degrade per its notes.
• Pairs with signal-scanner for TAM-scoped, scheduled, storage-backed scanning.