agentsclimarketplace
Sign in

Portco anomaly monitor

Skill loureiro-fernando/portfolio-pulse/skills/portco-anomaly-monitor

Trigger this skill when a portfolio company KPI (revenue, burn, churn_pct) arrives and you need to (1) detect if it's anomalous vs history and peer set, (2) gather context from internal docs and Slack, (3) classify severity per tenant policy, and (4) post a structured alert. Use when a user mentions a portco KPI drop/spike, says "investigate portco X", or asks for a portfolio health check. Do NOT trigger for forward-looking projections or pure reporting - only for actual incoming KPI events.From its SKILL.md

Install
npx -y skills add loureiro-fernando/portfolio-pulse --skill portco-anomaly-monitor

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

1.8 KB, 337 tokens by cl100k_base, as published. Nobody here has run it

portco-anomaly-monitor

Multi-tenant pipeline that detects, contextualizes, and classifies portfolio company KPI anomalies.

When to use

Trigger on incoming KPI events for portfolio companies. Example phrasings: "portco-1 just reported $3.1M revenue for April", "investigate the burn spike at GammaFin", "is BetaHealth's churn normal?".

Inputs (required)

  • tenant_slug (string) - e.g. "acme"
  • portco_id (string) - e.g. "portco-1"
  • metric (string) - one of: revenue, burn, churn_pct
  • value (float) - latest value
  • period (string) - YYYY-MM

What it does

  1. AnomalyDetector compares value vs 12-month history + sector peer average
  2. ContextBuilder enriches with PitchBook + Egnyte docs + Slack
  3. SeverityClassifier applies tenant policy -> JSON verdict
  4. Posts to Slack #portfolio-pulse if severity >= attention
  5. Persists AgentRun + Alert in Postgres
  6. Streams events to /stream/{tenant_slug} SSE

Outputs

  • alert_id (int) of created Alert
  • severity (info | attention | urgent)
  • requires_human (bool)
  • summary (string)

Cost

~$0.06 per invocation with Haiku 4.5.

See also

  • script.py for programmatic invocation
  • instructions.md for human operator workflow
  • evals/ for 5 deterministic test cases

What ships with it: 8 files

15.1 KB alongside SKILL.md, 7 of them executable

evals/

Keep looking

Skills are one crate of 326,286. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.