bluejay

Plugin bluejay from marketplace bluejay-skills in bluejay-ai-dev/bluejay-skills. Commands read from the repo's manifest — 8 days ago.

Skills (24)

• agent-onboard

  Take a brand-new Bluejay agent from "doesn't exist" to "exists and demonstrably answers a basic call" — create it with add_agent, then hand off to a smoke test. Use when the user wants to add a new agent and confirm it works, or invokes /bluejay:agent-onboard. (For first-run Bluejay setup as a whole

• agent-regression

  Re-test a Bluejay agent and tell the user whether a recent change made it better, worse, or held steady versus its last completed run — with the specific cases that regressed. Use when the user wants to verify a change didn't regress an agent, run a regression check, or invokes /bluejay:agent-regres

• agent-scorecard

  Produce a single health snapshot of a Bluejay agent across both testing (simulation pass rate + trend, weakest metrics) and production (call volume, status mix, failure rate), with the top things to fix. Use when the user wants an overall health read on an agent, a scorecard, or invokes /bluejay:age

• build-dashboard

  Design the right dashboard for watching a Bluejay agent's health — inspect the agent's real traffic and metrics, then hand the user a concrete, ready-to-build widget list (metric + chart type + filter) and a link to create it. Use when the user wants a dashboard of the metrics that matter for an age

• compare-runs

  Compare two Bluejay simulation runs to show exactly what changed — at the metric level and the case level (which cases flipped pass→fail and fail→pass). Use when the user wants to compare any two runs (e.g. before/after a prompt change), or invokes /bluejay:compare-runs. For an automatic "did my lat

• coverage-audit

  Map which branches of a Bluejay agent's conversation flow are exercised by existing test personas and which are blind spots, then propose personas to close the gaps. Use when the user wants to find untested paths, audit test coverage, or invokes /bluejay:coverage-audit.

• custom-metric-builder

  Convert a user's quality bar into well-formed Bluejay custom metrics — sharp, single-purpose pass/fail (or enum/quantitative) checks with clear judge prompts that produce scorable judgements on each call. Use when the user wants to create evals or metrics that grade what an agent should or shouldn't

• failed-call-triage

  Turn a pile of failed Bluejay production calls into a ranked list of failure modes the user can act on — clustered from real transcripts, with counts, examples, and prioritized fixes. Use when the user wants to know why live calls are failing and what to fix first, or invokes /bluejay:failed-call-tr

• knowledge-gap

  Turn "the agent fumbled some questions" into a concrete, prioritized list of knowledge-base additions, grounded in real failed simulation results and transcripts. Use when the user wants to find what the agent doesn't know and should, or invokes /bluejay:knowledge-gap.

• language-audit

  Show the language distribution of a Bluejay agent's calls and flag languages where the agent struggles (higher failure/short-call rates, mis-detection, fallback to English). Use when the user wants to see the language mix of calls and how well each is handled, or invokes /bluejay:language-audit.

• latency-triage

  Locate where a Bluejay agent's response time is being spent and which calls are worst, then name the most likely contributor (LLM, TTS first-audio, tool round-trips) and a concrete next step. Use when the agent feels slow and the user wants to find where the time goes, or invokes /bluejay:latency-tr

• load-test

  See how a voice agent holds up under many concurrent calls — dropped calls, latency spikes, error rates — by running a high-concurrency Bluejay simulation. Use when the user wants to load-test, stress-test capacity, or check behavior under concurrent calls, or invokes /bluejay:load-test.

• new-agent-smoke-test

  Run a quick end-to-end health check on a newly built or just-changed Bluejay agent — a small simulation with 3-5 sanity personas — to confirm it can hold a basic conversation before investing in bigger test suites. Use when the user wants to smoke-test or sanity-check an agent, or invokes /bluejay:n

• onboard

  First-run onboarding for Bluejay. Walks the user through creating and running their very first simulation entirely over the Bluejay MCP — connect (or register) their voice/chat agent, build a simulation whose digital humans match their real use case, queue the run, and hand them the live run link. E

• persona-suite

  Turn a short description of the callers you want to test into a diverse, runnable set of Bluejay digital humans — varied intent, sentiment, language, and edge cases — attached to a simulation. Use when the user wants to populate a simulation with realistic test callers or personas, or invokes /bluej

• production-pulse

  Give a fast, accurate read on live Bluejay production call traffic over a window — volume, inbound vs outbound, status mix, duration distribution, and notable shifts. Use when the user wants a quick pulse on production call activity, or invokes /bluejay:production-pulse. For failures use /bluejay:fa

• red-team-sweep

  Stress-test how a Bluejay agent handles hostile or manipulative callers — jailbreaks, prompt injection, PII extraction, social engineering, abuse, scope creep, multilingual evasion — then report concrete vulnerabilities and fixes. Use when the user wants to red-team, adversarially test, or probe an

• run-report

  Turn a Bluejay simulation run's raw results into an actionable report — verdict, per-metric table, and the failing calls with transcript links. Use when the user wants to understand how one run did in detail, "report on a run", or invokes /bluejay:run-report. For two-run comparisons use /bluejay:com

• run-simulation

  Start a Bluejay simulation against a voice agent right now and report the results. Resolves the agent and simulation, checks it has digital humans, queues the run, and gives a pass-rate readout when it completes. Use when the user wants to test an agent now, "run a sim", or invokes /bluejay:run-simu

• schedule-sim

  Put a Bluejay simulation on autopilot at a recurring cadence, or audit and edit what's already scheduled. Use when the user wants to run a test automatically on a schedule (nightly, weekly), review existing schedules, or invokes /bluejay:schedule-sim.

• self-improve

  Run an autonomous multi-round self-improvement loop for a voice/text agent against a Bluejay simulation. Generates pass/fail metrics from the agent's tools, runs the simulation, diagnoses failures from transcripts, fixes the system prompt, re-runs to verify each fix, and iterates until the target pa

• sms-simulation

  Test a text, SMS, or web-chat agent with a Bluejay simulation and report how the conversations went, using the correct queue tool for the channel. Use when the user wants to test an SMS/text/chat agent rather than a voice agent, or invokes /bluejay:sms-simulation.

• transcript-deep-dive

  Analyze one Bluejay call in depth — simulation result or production call — to explain not just that it failed but exactly where and why, and recommend one concrete fix. Use when the user wants to understand one specific call or test result in detail, or invokes /bluejay:transcript-deep-dive.

• workflow-branch

  Add a new branch (node + the edge that reaches it, like an escalation route) to a Bluejay agent's multi-node workflow graph, staging it on the Bluejay side and only pushing it live on explicit user approval. Use when the user wants to extend an agent's workflow with a new path or branch, or invokes

MCP servers (1)

• bluejay

  connects to: https://api.getbluejay.ai/mcp

Similar plugins