pm-agentops

Review an LLM agent design and find where it will be unreliable, expensive, or unsafe. Use when asked to review an agent architecture, critique a multi-step/tool-using agent, debug an agent that loops or goes off-task, or harden an agent before launch. Produces a structured review — task fit, control flow, tools, memory/context, failure handling, cost, and safety — with prioritised findings and fixes.

Run a blameless postmortem for an incident caused by an AI agent or LLM feature — hallucinated facts shipped to users, runaway tool use, prompt injection, cost blowouts, or wrong actions taken autonomously. Use when asked to write up an AI incident, analyse why an agent did something wrong, or produce corrective actions after an LLM failure. Produces a structured postmortem with trace reconstruction, a root-cause layer analysis, and corrective actions including a permanent regression case. For non-AI production incidents use incident-postmortem.

Specify the tracing, metrics, and alerting for an AI agent or LLM feature in production. Use when asked what to log for an LLM app, design agent tracing or spans, define quality and cost monitors, or answer 'how do we know if the agent is misbehaving?'. Produces an observability spec with a trace schema, metric definitions with owners and alert thresholds, sampling and retention policy, and a privacy note for logged content.

Design an evaluation plan for an LLM or AI feature before shipping it. Use when asked how to evaluate a prompt/model/agent, set up an eval harness, define quality metrics for an AI feature, or build a regression gate. Produces an eval plan — task definition, datasets, metrics & rubrics, baselines, automated + human evals, a pass bar, and a regression gate.

Review what an LLM feature or agent actually puts in its context window — and find what's bloating, missing, or fighting itself. Use when asked to review a system prompt and context assembly, cut token usage without losing quality, debug an agent that ignores instructions, or audit how retrieval results, history, and tool definitions are packed into the window. Produces a context inventory with a keep/cut/restructure verdict per component, ordering and caching fixes, and a token budget. For wording-level prompt tuning use prompt-optimizer.

Model the cost and latency of an LLM feature before it ships and surprises the bill. Use when asked to estimate LLM API costs, set a latency/token budget, decide which model tier to use, or bring down the cost of an AI feature. Produces a cost & latency budget — token math per request, monthly cost projection, model tiering, caching/streaming levers, p95 latency targets, and a guardrail/alert plan.

Plan the migration of an LLM feature from one model to another without breaking production. Use when a model is being deprecated, a newer model looks better or cheaper, or when asked how to upgrade models safely, run shadow traffic, or set rollback criteria for a model change. Produces a phased migration plan with eval gates, shadow/canary stages, prompt-adaptation notes, and rollback triggers. For choosing which model in the first place use model-selection-advisor.

Design a regression test suite that catches an LLM feature getting worse when the prompt, model, or context changes. Use when asked to stop prompt changes breaking production, set up golden tests or CI gates for an LLM feature, or test a model/prompt upgrade before shipping it. Produces a golden case set, per-case pass criteria, CI gate thresholds, and a triage protocol for failures. For designing first-time evaluation of a new feature use ai-eval-plan instead.