Inspect agent telemetry and operate a FailproofAI Cloud deployment via the fp CLI.
npx skills add https://github.com/FailproofAI/failproofai --skill fp-cloud-cli
Please help me install this Agent Skill. Skill: fp-cloud-cli Source: https://github.com/FailproofAI/failproofai/tree/main/fp-cloud-cli/skill Command: npx skills add https://github.com/FailproofAI/failproofai --skill fp-cloud-cli
Teams running production AI agents lack a quick way to answer "how are my agents doing?" — checking errors, sessions, eval scores, and usage — and to operate their agent-observability deployment (issues, audits, API keys, users, policies) without opening a dashboard.
fp --json errors --since 24h --aggregate, drills into failing sessions, and reports the answer in prose.Ask the assistant to check whether any of your production agents errored in the last 24 hours using the fp CLI.
High-intent search queries and answers about installing and using this skill.
Run fp --json errors --since 24h --aggregate for window totals, then fp --json errors --since 24h --all --limit 1000 to break failures down by session and error type. Use fp --json sessions --status error to list failed runs.
Run fp --json events --session-id <id> --all --limit 1000 for the run's timeline. Add --full to include raw payloads, but always bound it to one session since the full feed is slow at scale.
Yes, pass --api-key or set FP_API_KEY for non-interactive contexts. Note that login, orgs, agent, policies, fleet, and guardrails commands are session-only and exit with code 2 under a key, and you should pass --org explicitly with keys.
The --limit flag defaults to 50 and --all paginates only up to that limit. Pass a high explicit cap like --all --limit 1000 for a full sweep, or use --aggregate for whole-window totals.
Run fp audits run <name> to queue a run, poll fp audits runs <name> until it succeeds, then list findings with fp audits findings --audit <name>. Triage with audits ack, mute, dismiss, or resolve, which stay in sync with the linked issue.
Do not use it for writing evaluator scoring logic, adding SDK instrumentation to your app, debugging the collector daemon, or unrelated dev work like CI failures. It only covers operating and observing a FailproofAI Cloud deployment.
grafanaCollect continuous runtime profiles from applications to identify performance hotspots.
multica-aiGuides agents through production incident triage, deploy correlation, and rollback requests.
aeoncity-demoMonitor skill run counts and alert on daily limit breaches.
MoonAxisQuery Azure Monitor metrics, Application Insights traces, and Log Analytics logs via KQL.