mas-ai-workflows

Plugin mas-ai-workflows from marketplace major-ai-skills in alivirgo/Major-AI-Skills. Commands read from the repo's manifest — yesterday.

Skills (10)

• agent-injection-boundary-test

  Test authorized agents against indirect prompt injection in retrieved docs and tool outputs using paired clean/poison fixtures, synthetic canaries, and sandboxed tools. Evidence for these cases—not a claim injection is solved.

• agent-tool-replay-test

  Replay recorded agent tool/LLM trajectories against deterministic fixtures to test argument validation, errors, idempotency, and side-effect boundaries—separate from live tool-selection quality.

• ai-citation-verification

  Verify AI citations across structural schema, resolvability (registry/URL/DOI), and semantic support of each claim by the cited span—not the whole context. Use for RAG answers, research agents, and regulated audit trails.

• ai-evaluation-dataset

  Build versioned JSONL evaluation datasets for AI workflows with golden/regression/holdout layers, leakage checks, rubrics, and provenance hashes. Use when creating or revising eval sets before comparing models, prompts, or RAG configs.

• ai-human-handoff-contract

  Define and test AI→human handoff as a versioned state machine with multi-signal triggers, review packets, approval binding, and no silent consent. Use for escalations, approvals, and irreversible agent actions.

• ai-pii-redaction-review

  Privacy-review AI payloads with field-aware redaction, typed stable placeholders, leakage scans across prompts/logs/tools/RAG, and residual-risk reporting before external submission. Use before sending data to cloud LLMs or sharing traces.

• llm-cost-latency-benchmark

  Benchmark AI workflow cost and latency with TTFT/TPOT/ITL/e2e percentiles, cache-aware token billing, goodput under SLOs, and a quality floor. Use before claiming a model, prompt, or infra change is cheaper or faster.

• llm-json-contract-check

  Validate AI-generated JSON against schema and business rules, separating refusals, truncation, parse failures, schema violations, and domain invariant breaches. Use before any tool call, DB write, or external side effect consumes model output.

• prompt-regression-gate

  Gate prompt, model, tool-schema, or packer changes with paired runs on a frozen eval set, slice-level regression checks, relative baselines, and explicit go/no-go thresholds. Use before merging prompt PRs or swapping models.

• rag-retrieval-audit

  Diagnose missing evidence in a RAG pipeline with labeled queries, chunk/trace inspection, Recall@K/MRR/nDCG, and stage-isolated failure taxonomy. Use when retrieval looks fine but answers fail, or when index/chunk/filter/rerank changes need before/after proof.

Similar plugins