mustflow 2.116.3 → 2.117.0

package/templates/default/locales/en/.mustflow/skills/INDEX.md

@@ -2,7 +2,7 @@
 mustflow_doc: skills.index
 locale: en
 canonical: true
-revision: 277
+revision: 289
 authority: router
 lifecycle: mustflow-owned
 ---
@@ -50,6 +50,66 @@ refer to `AGENTS.md` and `.mustflow/config/commands.toml` to implement the most
   current deliverable, the task has both a material uncertainty signal and a material consequence
   signal, and no narrower primary route owns the complete problem. Before implementation, switch to
   the narrowest matching implementation skill.
+- Use `automation-investment-case-review` as a primary route when deciding whether an AI agent,
+  workflow automation, internal tool, or integration is worth building, expanding, replacing, or
+  retiring from accepted-outcome cost, human comparison, supervision, failure recovery,
+  maintenance, break-even, throughput value, effective lifetime, NPV, and independent safety gates.
+- Use `product-onboarding-activation-review` as a primary route when account-first versus pre-account
+  core value, signup checkpoints, guest-to-account transfer, authentication choices, first-run
+  questions, personalization, editable samples, contextual help, tutorial depth, progressive
+  disclosure, or a focused primary action must improve a user-owned activation result, retained use,
+  or paid conversion without excluding account or onboarding abandoners from the denominator.
+- Use `subscription-retention-profit-review` as a primary route when cancellation reasons, save
+  offers, downgrade, pause, credits, discounts, dormant-user reactivation, win-back, offer ordering,
+  or retained-value claims must be decided from holdout-backed incremental contribution rather than
+  offer acceptance, natural return, or paused-user counts.
+- Use `product-engagement-retention-review` as a primary route when weekly value reports, saved-time
+  summaries, artifact recaps, streaks, grace or recovery, attendance rewards, abandoned-work
+  reminders, or notification timing must improve completed value and retained use without counting
+  activity theater, reward claims, natural return, or message opens as durable retention.
+- Use `credit-monetization-integrity-review` as a primary route when credit-offer timing, discount
+  versus bonus framing, pack count or spacing, purchased-credit expiry, subscription rollover,
+  balance-class rights, spend order, or exact or bounded pre-execution pricing must improve
+  long-horizon contribution without deceptive equivalence, breakage theater, or surprise deductions.
+- Use `pricing-model-integrity-review` as a primary route when trial card collection, lifetime or
+  founder access, local currency, regional pricing, subscription versus usage or hybrid pricing,
+  price elasticity, annual prepay, or annual discounts must improve eligible-cohort contribution
+  without hidden renewal, unbounded future-cost rights, or cash-revenue confusion.
+- Use `game-economy-monetization-review` as a primary route when paid or rewarded-ad revives, lives,
+  energy, refills, unlimited play, VIP benefits, IAP cannibalization, or content pacing must improve
+  long-horizon player contribution without selling ranked outcomes or erasing meaningful failure.
+- Use `freemium-ad-monetization-review` as a primary route when free-tier limits, ad-supported access,
+  interstitial or rewarded-ad placement, first-value suppression, premium ad removal, or ad
+  frequency must improve contribution without interrupting active work or holding results hostage.
+- Use `referral-incentive-integrity-review` as a primary route when inviter, invitee, dual-sided, or
+  tiered referral rewards, attribution, valid-referral qualification, vesting, reversal, self-referral,
+  reward farming, or incremental referral contribution is created, changed, reviewed, or reported.
+- Use `llm-product-monetization-review` as a primary route when token versus task pricing, bounded AI
+  work units, standard versus precision tiers, free versus paid queue priority, automatic failure
+  restoration or retry rights, model disclosure or pinning, managed provider cost, BYOK, customer API
+  keys, platform fees, or LLM failed-work charging changes.
+- Use `game-liveops-commerce-integrity-review` as a primary route when season passes, memberships,
+  liveops cadence, cosmetics, power goods, deterministic bundles, loot boxes, odds, pity, payer
+  concentration, or game commerce production and fairness risk changes.
+- Use `growth-distribution-integrity-review` as a primary route when result watermarking, embedded
+  attribution, affiliate or influencer compensation, recurring commission, partner attribution,
+  owned-product cross-promotion, portfolio fatigue, brand confusion, or incremental distribution
+  contribution changes.
+- Use `product-portfolio-integrity-review` as a primary route when unified accounts, shared credit
+  wallets, individual versus bundled subscriptions, portfolio entitlements, umbrella or endorsed
+  brands, cross-product reuse, cannibalization, or failure spillover changes.
+- Use `localization-market-expansion-review` as a primary route when first-market language,
+  simultaneous versus staged locale rollout, AI translation review depth, exploratory versus
+  revenue-ready locales, localized SEO demand, or per-locale retained contribution changes.
+- Use `desktop-commercial-distribution-review` as a primary route when a desktop product compares
+  website delivery with Microsoft Store, Mac App Store, or hybrid distribution across commerce,
+  licensing, trust, capabilities, updates, fees, conversion, and retained contribution.
+- Use `client-platform-strategy-review` as a primary route when a product compares web-first,
+  PWA, cross-platform, or native clients, calculates native break-even, or chooses local-first,
+  cloud-first, offline, sync, privacy, and recovery authority.
+- Use `service-portfolio-capital-allocation-review` as a primary route when shared backend investment,
+  correlated failure cells, service continuation or shutdown, rescue experiments, or founder time
+  between new launches and existing-product improvements changes.
 - Use `technology-stack-selection` as the narrower primary route when the decision chooses,
   adopts, replaces, rejects, or standardizes a technology stack, vendor, framework, database,
   queue, auth, payment, AI provider, hosting, deployment, build, ORM, or observability surface.
@@ -84,10 +144,35 @@ refer to `AGENTS.md` and `.mustflow/config/commands.toml` to implement the most
   token, first useful output, streaming, output length, LLM round trips, tool wait, prompt-cache
   latency, model routing speed, speculative or parallel work, realtime continuation, priority
   tiers, or predicted outputs need latency review.
+- Use `llm-model-routing-integrity-review` as a primary route when model selection, small-to-large
+  cascades, escalation, fallback, stage switching, verifier thresholds, context-handoff loss,
+  distribution shift, or OOD behavior must preserve accepted outcomes while minimizing total route,
+  verification, repair, latency, and operating cost.
 - Use `agent-execution-control-review` as a primary route when autonomous or semi-autonomous LLM
   agents, agentic workflows, planners, executors, verifiers, tool contracts, tool-call gates,
-  approval or interrupt flows, durable agent state, handoffs, guardrails, loop budgets, retry
-  classification, trace evaluation, or outcome metrics need execution-control review.
+  risk-tiered approval or interrupt flows, reversibility, external effects, durable agent state,
+  handoffs, guardrails, loop budgets, retry classification, trace evaluation, or outcome metrics
+  need execution-control review.
+- Use `agent-planning-recovery-review` as a primary route when a long-running agent needs a stable
+  global goal and dependency graph, short rolling plans, event-triggered replanning, append-only
+  state, rebuildable snapshots and prompt views, deterministic replay, or effect identity across
+  interruptions and replans.
+- Use `agent-release-bundle-rollout-review` as a primary route when models, rendered prompts, tools,
+  policies, retrieval, memory, runtimes, and evaluators must form one immutable agent behavior
+  bundle promoted through offline replay, effect-free shadowing, sticky canaries, staged expansion,
+  hard safety gates, compatibility, rollback, and attenuation-only emergency restriction.
+- Use `agent-runtime-multi-worker-review` as a primary route when a production agent runtime must
+  decide whether independent parallel work justifies several agents, then define orchestrator-worker
+  topology, specialized roles, artifact ownership, correlation controls, bounded fan-out, and one
+  central verifier.
+- Use `agent-runtime-isolation-review` as a primary route when durable agent state must outlive
+  short-lived queued workers, model-directed shell, browser, file, package, or code work needs an
+  ephemeral sandbox, credentials must be temporary and task-scoped, or production effects need a
+  trusted policy broker outside the sandbox.
+- Use `agent-memory-context-governance-review` as a primary route when persistent user or project
+  memory, conversation history, memory extraction or consolidation, summaries, structured task
+  state, long-context assembly, retrieval admission, supersession, expiry, deletion, privacy, or
+  memory quality needs scope, provenance, freshness, authority, and lifecycle review.
 - Use `browser-automation-reliability-review` as a primary route when browser automation,
   UI automation, Playwright, Selenium, Puppeteer, WebDriver, browser-driving agents, flaky
   selectors, page readiness, auth state, CAPTCHA or anti-bot handling, rate limits, screenshot
@@ -129,11 +214,12 @@ refer to `AGENTS.md` and `.mustflow/config/commands.toml` to implement the most
   review for idempotency, ordering, ownership, amount, currency, retry, reconciliation, or audit
   risk.
 - Use `credit-ledger-integrity-review` as an adjunct when credits, points, wallet balances, reward
-  points, prepaid or usage credits, balance deductions, accruals, refunds, expirations,
+  points, prepaid or usage credits, purchased versus subscription or promotional balance classes,
+  balance deductions, accruals, refunds, expirations, rollovers, user-visible price quotes,
   reservations, captures, releases, admin adjustments, ledger tables, balance caches, or
   reconciliation jobs need ledger integrity review for idempotency, atomic balance changes,
-  concurrency, ordering, ownership, amount precision, policy snapshots, expiry lots, audit, or
-  reconciliation risk.
+  concurrency, ordering, ownership, amount precision, policy snapshots, quote binding, expiry lots,
+  audit, or reconciliation risk.
 - Use `dual-write-consistency` as an adjunct when one logical operation commits local state and
   separately publishes, projects, calls, or writes another independently committed system, so
   crash points, outbox or inbox delivery, reconciliation, and eventual convergence must be proved.
@@ -444,6 +530,7 @@ refer to `AGENTS.md` and `.mustflow/config/commands.toml` to implement the most
 - Use `retry-policy-integrity-review` as an adjunct when retry loops, SDK or client retry configs,
   backoff, jitter, timeout, deadline, `Retry-After`, retry predicates, layered retries, circuit
   breakers, bulkheads, token buckets, queue redelivery, broker retries, cancellation-aware sleeps,
+  prompt repair, model or tool fallback, replanning, repeated failure signatures, self-correction,
   or retry observability can amplify failures, duplicate side effects, hide permanent errors,
   exhaust pools, or overload dependencies.
 - Use `queue-processing-integrity-review` as an adjunct when queues, streams, pub/sub handlers,
@@ -580,16 +667,22 @@ routes. Event routes stay inactive until their event occurs.
 | LLM answers, RAG responses, citations, source grounding, claim extraction, evidence IDs, answerability states, abstain behavior, retrieval thresholds, tool-backed facts, output validators, LLM judges, or hallucination-control metrics are created, changed, reviewed, or reported | `.mustflow/skills/llm-hallucination-control-review/SKILL.md` | Answer contract ledger, evidence ledger, claim ledger, tool ledger, validator ledger, eval ledger, observability ledger, changed files, and command contract entries | Answerability states, abstain states, missing-information states, source-coverage gates, claim maps, evidence-ID requirements, citation validators, retrieval thresholds, chunk metadata, tool-parameter ownership, deterministic calculators, domain validators, eval fixtures, tests, docs, route metadata, and directly synchronized templates | unsupported factual claim, fabricated citation, source ID invention, weak retrieval gate, noisy semantic-only retrieval, chunk context loss, summary-on-summary hallucination, guessed tool parameter, model arithmetic, source-priority conflict, LLM judge overtrust, low-temperature theater, missing abstain path, missing dirty eval, false citation metric gap, or unobservable grounding drift | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Hallucination-control surface reviewed, answerability and abstain states, evidence IDs, claim map, citations, source coverage, validators, retrieval thresholds, tool ownership, evals, metrics, verification, and remaining hallucination-control risk |
 | RAG, knowledge-base answer, grounded chat, citation answer, retrieval-augmented support bot, or document QA flow is wrong, stale, unsupported, slow, leaking data, over-refusing, or not yet localized to ingestion, parsing, document metadata, frontmatter schema, source maps, heading paths, chunking, retrieval, filtering, reranking, context assembly, prompt construction, generation, citation validation, or answerability boundaries | `.mustflow/skills/rag-pipeline-triage/SKILL.md` | Symptom classification, trace ledger, source ledger with original/index/prompt text, stable source/doc/chunk ids, document type/status/authority, routing summaries, document graph links, comparison ledger, eval ledger, privacy ledger, changed files, and command contract entries | End-to-end trace preservation, source availability and parsed-text checks, metadata, source maps, ACL prefilters, heading paths, semantic code chunks, table record shape, chunk graph, source-of-truth and supersession checks, duplicate and stale source checks, no-retrieval/current-context/gold-context comparison, keyword/vector/hybrid/retriever/reranker/context/prompt/generation/citation/answerability localization, safe synthetic fixtures, metrics, docs, and directly synchronized templates | model scapegoating, tuning top-k before source proof, original-document theater while parsed text is broken, generic heading coordinates, generated summary cited as source, missing source map, source/doc/chunk id collapse, stale source mixing, draft or meeting-note evidence outranking canonical documents, ACL inherited after retrieval, filter blamed as vector failure, fixed line-count code slicing, table number without row or column meaning, reranker candidate starvation, critical evidence truncated or buried, retrieved text treated as instruction, citation decoration, answerability missing state, private corpus dump, access filter bypass, or single satisfaction score hiding layer failure | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | RAG pipeline triaged, localized boundary, trace/source/comparison/eval/metric/privacy ledgers, metadata/frontmatter/source-map/ACL/chunk-graph/document-graph findings, fix or recommendation, evidence level, verification, and remaining RAG pipeline risk |
 | LLM API calls, prompt assembly, chat history, RAG context, document metadata, chunk summaries, prompt packing, question-scoped compression, evidence cards, tool schemas, structured output schemas, model routing, reasoning settings, token budgets, provider prompt caching, app-level response caching, retries, batch or flex processing, predicted outputs, image or file inputs, or LLM cost metrics are created, changed, reviewed, or reported | `.mustflow/skills/llm-token-cost-control-review/SKILL.md` | Cost surface ledger, request ledger with authority lanes, cache ledger, context ledger with inclusion tests, block role tags, filter-only versus LLM-visible metadata, source maps, state cards, and evidence cards, output ledger, routing ledger, observability ledger, changed files, and command contract entries | Request builders, prompt prefix ordering, canonical serialization, prompt hashes, cache keys, token counters, budget guards, model routers, context trimming, RAG packing, question-scoped compression, slot records, state snapshots, original/index/prompt text separation, source-map references, tool and schema payloads, output patch formats, retry repair paths, metrics, logs, tests, docs, route metadata, and directly synchronized templates | prompt-cache prefix drift, volatile field before stable prefix, unmeasured token count, full transcript replay, state delta without baseline, answer-irrelevant context hoarding, RAG chunk bloat, generated summary treated as original evidence, search-only metadata treated as answer text, missing source coordinate, missing token_count, compression not evaluated, oversized tool or JSON schema payload, expensive model default, unbounded reasoning, no visible output after token cap, full-output regeneration, full-context retry replay, app cache key leak, predicted-output cost confusion, image or file token surprise, or per-call cost hiding cost-per-success regression | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `prompt_cache_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | LLM token-cost surface reviewed, cost unit and measurement source, stable prefix and cache behavior, authority lanes, app cache/history/RAG/metadata/source-map/prompt-packing/evidence-card/tool/schema/input choices, routing/reasoning/output/retry/Batch/Flex/prediction choices, observability, verification, and remaining token-cost risk |
+| LLM requests are selected, cascaded, escalated, or switched among models by task, stage, confidence, verifier result, distribution shift, cost, latency, quality, safety, or context handoff | `.mustflow/skills/llm-model-routing-integrity-review/SKILL.md` | Accepted-outcome contract, capable-model baseline, route and evidence ledgers, total cost and latency ledger, distribution and OOD evidence, context-handoff contract, hard-safety routes, and command contract entries | Model router, route features and versions, deterministic pre-routing, external verifier, escalation and abstention, OOD fallback, context handoff, retry ownership, shadow evaluation, metrics, tests, docs, route metadata, and directly synchronized templates | per-call price presented as savings, self-confidence treated as probability, cheap-first route without a false-accept gate, context loss across model switches, route retry budget reset, OOD traffic entering an unsupported route, hard-safety decision averaged into utility, or universal threshold copied from a benchmark | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Model-routing boundary reviewed, accepted-outcome baseline, external calibration, total cost per accepted outcome, route and handoff decisions, OOD and hard-safety fallback, verification, and remaining model-routing risk |
 | LLM response latency, time to first token, first useful output, streaming, output length, LLM round trips, tool-call wait, prompt-cache latency, model routing, speculative or parallel execution, realtime continuation, priority tiers, predicted outputs, or user-perceived AI speed are created, changed, reviewed, or reported | `.mustflow/skills/llm-response-latency-review/SKILL.md` | Latency target ledger, request timeline ledger, call graph ledger, output ledger, cache ledger, routing ledger, observability ledger, changed files, and command contract entries | Streaming paths, first-useful-output contracts, request timeline metrics, call graph simplification, parallel or speculative work, model routers, fallback cascades, output caps, schema shortening, prompt-cache prefix ordering, cache keys, realtime continuation, priority-tier routing, timeout and cancellation behavior, tests, docs, route metadata, and directly synchronized templates | slow first token, useless streamed preamble, extra sequential LLM round trip, tool wait blocking first output, cache-prefix drift, unmeasured cache miss, verbose output drift, long JSON key or enum overhead, RAG chunk bloat, router escalation loop, prediction-token mismatch, priority-tier misuse, unsafe speculative work, missing cancellation, or raw prompt telemetry leak | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | LLM response-latency surface reviewed, latency unit and request timeline, round trips, parallel/tool/stream/cancel behavior, output/schema/cache/history/RAG/routing/fallback/prediction/realtime/priority choices, observability, verification, and remaining response-latency risk |
-| Autonomous or semi-autonomous LLM agents, agentic workflows, planners, executors, verifiers, tool contracts, tool-call gates, human approval or interrupt flows, durable agent state, handoffs, guardrails, loop budgets, retry policies, trace evaluation, or agent outcome metrics are created, changed, reviewed, or reported | `.mustflow/skills/agent-execution-control-review/SKILL.md` | Autonomy ledger, stage gate ledger, role separation ledger, tool contract ledger, effect ledger, state and resume ledger, memory and context ledger, handoff and guardrail ledger, loop, retry, and budget ledger, trace and eval outcome ledger, changed files, and command contract entries | Workflow-versus-agent routing, stage gates, planner/executor/verifier boundaries, tool contracts, tool argument ownership, draft/execute separation, idempotency keys, approval records, durable checkpoints, state schema versions, memory partitions, handoff filters, guardrails, loop budgets, retry classification, trace spans, eval fixtures, tests, docs, route metadata, and directly synchronized templates | unnecessary autonomous agent, one model self-certifying success, ungated bad plan, ambiguous tool contract, guessed tool argument, relative-path trap, external effect before approval, missing idempotency key, interrupt replay side effect, stale state schema, over-shared handoff, misplaced guardrail, repeated tool loop, blind retry, final-answer-only eval, or unsafe trace data | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Agent execution-control surface reviewed, workflow-versus-agent decision, autonomy envelope, stage gates, role separation, tool contracts, approval and side-effect replay safety, state/resume/memory/handoff/guardrail/loop/retry/trace/eval checks, verification, and remaining agent execution-control risk |
+| Autonomous or semi-autonomous LLM agents, agentic workflows, planners, executors, verifiers, tool contracts, tool-call gates, risk-tiered approval or interrupt flows, reversible or irreversible effects, durable agent state, handoffs, guardrails, loop budgets, retry policies, trace evaluation, or agent outcome metrics are created, changed, reviewed, or reported | `.mustflow/skills/agent-execution-control-review/SKILL.md` | Autonomy, stage gate, role separation, tool contract, capability, effect, approval and reversibility, state and resume, memory and context, handoff and guardrail, loop, retry and budget, trace and eval outcome ledgers, changed files, and command contract entries | Workflow-versus-agent routing, semantic action grouping, forced-approval conditions, risk tiers, reversibility classes, scoped policies, just-in-time approval binding, delayed commit, exact rollback or compensation, planner/executor/verifier boundaries, tool contracts, idempotency and reconciliation, independent postcondition checks, durable checkpoints, guardrails, budgets, traces, eval fixtures, tests, docs, route metadata, and directly synchronized templates | high-impact action hidden by an average score, approval per tool call, approval before targets are known, compensation sold as rollback, plan drift reusing approval, unbounded standing permission, external effect before approval, generic Allow view, missing idempotency key, unknown-outcome replay, unsafe automatic recovery, one model self-certifying success, repeated tool loop, or unsafe trace data | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Agent execution-control surface reviewed, autonomy and semantic action boundary, forced-approval and risk decision, reversibility, policy and per-action approval binding, side-effect verification and recovery, state/resume/memory/handoff/guardrail/loop/retry/trace/eval checks, verification, and remaining agent execution-control risk |
+| A production LLM-agent runtime may use several agents or workers and must decide whether independent parallelism justifies its coordination cost, then define orchestrator-worker topology, specialized roles, artifact ownership, correlated-error controls, bounded fan-out, and central verification | `.mustflow/skills/agent-runtime-multi-worker-review/SKILL.md` | Independently verifiable work units, comparable single-agent baseline, benefit and coordination ledgers, topology and role ledgers, artifact and effect ownership, correlation ledger, completion policy, and command contract entries | Worker admission, star topology, delegation schema, role and capability boundaries, durable artifact handoff, central verifier, fan-out and depth budgets, cancellation, metrics, tests, docs, route metadata, and directly synchronized templates | multi-agent as default, arbitrary prompt-turn split, all-to-all transcript relay, duplicated correlated workers presented as independent experts, majority vote without an oracle, several owners for one artifact or effect, verifier bottleneck, unbounded fan-out, or gains caused only by extra compute | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Runtime multi-worker decision reviewed, single-agent baseline, independent work-unit and role evidence, topology and ownership, correlation and central-verification controls, budgets, verification, and remaining multi-worker risk |
+| Durable agent state must outlive short-lived execution while queued workers, ephemeral sandboxes, scoped mounts, temporary credentials, capability-specific queues, or a trusted effect broker constrain runtime blast radius | `.mustflow/skills/agent-runtime-isolation-review/SKILL.md` | Durable-state and execution-lifetime ledgers, control-plane authority, capability queue contract, sandbox manifest, credential and network policy, artifact admission, effect broker and reconciliation contract, maximum tolerable loss, and command contract entries | Event and checkpoint stores, leases and fencing, queue admission, ephemeral sandbox lifecycle, read-only input mounts, output staging, short-lived credentials, network deny rules, artifact validation, brokered effects, replay and recovery, tests, docs, route metadata, and directly synchronized templates | durable truth inside a worker filesystem or prompt, long-lived privileged worker, shared mutable workspace or credentials, generic queue with ambient capability, unrestricted egress, sandbox output admitted without validation, worker direct production access, timeout treated as failed effect, or isolation selected only by infrastructure price | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Runtime isolation boundary reviewed, control and execution plane split, state survival, worker and credential lifetime, capability queue and sandbox contract, artifact and effect admission, replay safety, verification, and remaining isolation risk |
+| A long-running LLM agent or agentic workflow needs a stable global goal, constraints, milestones, dependency DAG, irreversible checkpoints, short rolling plans, event-triggered replanning, compact prompt reconstruction, deterministic replay, or safe recovery across interruptions and UNKNOWN effects | `.mustflow/skills/agent-planning-recovery-review/SKILL.md` | Global contract, rolling-plan ledger, event envelope, reducer and projection contract, snapshot cursor, prompt-view contract, workflow/plan/step/effect/attempt identities, recovery and reconciliation owner, and configured command intents | Global and rolling plan schemas, milestone DAGs, replan triggers, append-only events, deterministic reducers, rebuildable projections and snapshots, compact prompt views, stable effect identity, reconciliation, outbox boundaries, compensation effects, tests, docs, route metadata, and synchronized templates | frozen detailed plan, replan-every-step drift, prompt or summary as authority, model-dependent replay, snapshot as irreplaceable truth, new effect ID after replan, timeout stored as failure, duplicate compensation, sequence-gap resume, stale owner, or raw log stuffed into context | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Global contract and horizon, replan triggers, event/projection/snapshot/prompt-view contract, identity and reconciliation findings, recovery evidence, verification, and remaining agent-planning recovery risk |
+| An LLM agent's model, settings, rendered prompts, examples, tool schemas, adapters, policy, retrieval, memory, runtime, evaluator, or eval dataset changes and must be released as one attributable immutable bundle through offline replay, shadow, canary, staged promotion, rollback, or emergency restriction | `.mustflow/skills/agent-release-bundle-rollout-review/SKILL.md` | Bundle manifest and digest, component versions, candidate/stable channels, assignment rule, sticky cohort, rollout and hard-gate ledger, effect suppression, compatibility matrix, rollback target, capability revocation, reconciliation and compensation owner, and configured command intents | Immutable behavior manifests, channel records, assignment and cohort rules, static/offline gates, write-free shadow adapters, canary and staged promotion gates, absolute and control comparisons, attenuation-only safety overlays, compatibility, rollback, observability, tests, docs, route metadata, and synchronized templates | moving alias in flight, candidate auto-promoted to stable, shadow with write tools, per-request random canary, easy-traffic-only cohort, aggregate quality hiding safety failure, relative-only gate, pointer-only rollback, incompatible old state, unrevoked capability, or unowned UNKNOWN effect | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Bundle identity, candidate/stable and cohort decision, offline/shadow/canary/promotion evidence, hard gates, compatibility, rollback/revocation/reconciliation readiness, verification, and remaining agent-release rollout risk |
 | Browser automation, UI automation, Playwright, Selenium, Puppeteer, WebDriver, computer-use or browser-driving agents, visual browser verification, flaky selectors, page readiness, authentication state, CAPTCHA or anti-bot handling, rate limits, screenshot checks, retry, timeout, human approval, or browser automation observability is created, changed, reviewed, triaged, or reported | `.mustflow/skills/browser-automation-reliability-review/SKILL.md` | Automation intent ledger, state ledger, readiness ledger, selector and action ledger, auth and identity ledger, external pressure ledger, verification ledger, agent and approval ledger, changed files, and command contract entries | Browser automation state machines, locator contracts, test IDs, accessible names, readiness assertions, frame and popup handlers, input verification, auth fixtures, per-worker account isolation, retry classification, timeout hierarchy, idempotency checks, rate-limit handling, approval gates, manual fallback states, traces, screenshots, redaction, cleanup, tests, docs, route metadata, and directly synchronized templates | sleep-as-readiness, `networkidle` faith, flaky selector string patch, hidden duplicate DOM, skeleton-as-content, stale element handle, force-click default, shared mutable account, CAPTCHA bypass, anti-bot evasion, retry storm, non-idempotent replay, timeout layer mismatch, screenshot-as-business-proof, unredacted trace data, page prompt injection, stale approval resume, coordinate click drift, or unverified browser success claim | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Browser automation surface reviewed, browser-versus-API boundary, automation owner, state/readiness/locator/actionability/auth/rate-limit/retry/timeout/idempotency decisions, screenshot and business-success evidence, agent page-content trust, approval and resume checks, verification, and remaining browser automation reliability risk |
-| Agent evaluation loops, trace or trajectory grading, LLM judges, verifier agents, outcome scoring, tool-call prechecks or postchecks, eval datasets, golden or dirty sets, pass@k or pass^k metrics, shadow environments, production-monitoring-to-eval pipelines, or agent regression gates are created, changed, reviewed, or reported | `.mustflow/skills/agent-eval-integrity-review/SKILL.md` | Outcome ledger, trace ledger, oracle ledger, tool-boundary ledger, dataset ledger, metric ledger, environment ledger, monitoring ledger, privacy ledger, changed files, and command contract entries | Outcome oracles, trace schemas, trajectory graders, deterministic checkers, model-judge rubrics, human-review sampling, tool prechecks and postchecks, tool-result evidence packets, eval fixtures, golden and dirty sets, shadow-environment adapters, monitoring-to-eval candidate flows, tests, docs, route metadata, and directly synchronized templates | final-answer-only scoring, LLM judge as sole oracle, reasoning claim treated as evidence, self-reflection certifying success, missing final environment state, ungraded unsafe trajectory, missing tool precheck or postcheck, brittle exact tool-order assertion, uncalibrated judge drift, pass@k masking unreliable pass^k, dirty set gating flakiness, raw trace data leak, or production failure not entering evals | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Agent eval-integrity surface reviewed, outcome oracle, final-state and trajectory checks, deterministic/model/human oracle split, tool boundary evidence, golden/dirty/capability/regression metrics, shadow environment, monitoring-to-eval loop, trace privacy, verification, and remaining agent eval-integrity risk |
+| Agent evaluation loops, deterministic gates, trace or trajectory grading, LLM judges, verifier agents, repair cascades, outcome scoring, tool-call prechecks or postchecks, fixed, replay, or generated eval datasets, pass@k or pass^k metrics, shadow environments, production-monitoring-to-eval pipelines, or agent regression gates are created, changed, reviewed, or reported | `.mustflow/skills/agent-eval-integrity-review/SKILL.md` | Outcome, decision-loss, trace, oracle independence, tool-boundary, dataset lifecycle, metric, environment, monitoring, and privacy ledgers, changed files, and command contract entries | Deterministic-first asymmetric cascade, evidence-backed repair, semantic verification, bounded adjudication, pass/fail/hold states, outcome and trajectory oracles, fixed regression, recent shadow replay, generated exploration, contamination and expiry controls, operational metric families, alerts, tests, docs, route metadata, and directly synchronized templates | final-answer-only scoring, false rejection without reproducible evidence, correlated judges treated as independent, self-reflection certifying success, ambiguous verdict forced to pass or fail, one easy set masking another regression, stale expected truth, train-eval leakage, unsafe trajectory, pass@k masking unreliable pass^k, aggregate metrics hiding a safety incident, raw trace leak, or production failure not entering evals | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Agent eval-integrity surface reviewed, decision-loss and asymmetric cascade, final-state and trajectory checks, deterministic/verifier/adjudicator/human oracle split, dataset roles and lifecycle, contamination, operational metrics and alerts, shadow environment, monitoring-to-eval loop, trace privacy, verification, and remaining agent eval-integrity risk |
+| Persistent LLM-agent or assistant memory, user preferences, project decisions, conversation history, memory extraction or consolidation, summaries, structured task state, long-context assembly, memory retrieval or admission, supersession, expiry, deletion, user controls, or memory-quality metrics are created, changed, reviewed, or reported | `.mustflow/skills/agent-memory-context-governance-review/SKILL.md` | Memory-class, authority, record, lifecycle, context-assembly, retrieval, privacy and deletion, and eval ledgers plus command contract entries | Memory schemas and classes, extraction gates, version and supersession links, lifecycle states, freshness and validity, retrieval filters and admission, context layers, summary regeneration, long-context fallback, tombstones, deletion propagation, user controls, eval fixtures, tests, docs, route metadata, and directly synchronized templates | generic memory record, model inference stored as fact, current instruction overridden by stale memory, summary-on-summary drift, full history as default context, similarity before access filters, cross-user or cross-project leakage, role-play contamination, retrieved memory treated as command, sensitive or secret retention, partial deletion claimed complete, tombstone missing, recall-only quality claim, or unverified production erasure | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Agent memory and context-governance surface reviewed, memory classes and authority, lifecycle and record contract, context and retrieval admission, privacy and deletion, memory-quality evidence, verification, and remaining memory, context, privacy, deletion, or production-evidence risk |
 | Code review or implementation needs abstraction-boundary triage for deciding whether a helper, interface, adapter, service, policy object, strategy, mapper, DTO, base class, component, or shared module should exist, stay duplicated, be inlined, move side effects out of the core path, preserve layer import contracts, or delete generated boilerplate after refactoring | `.mustflow/skills/abstraction-boundary-review/SKILL.md` | User goal, current diff or target files, change-reason ledger, future-change scenarios, domain vocabulary, ownership map, candidate boundary ledger, public-promise ledger, side-effect ledger, test-shape ledger, layer-contract ledger, layer-integrity ledger, type-existence ledger, public-contract ledger, existing helpers or patterns, and configured command intents | Local abstractions, deliberate duplication, policy owners, adapters, mappers, result or variant types, side-effect shells, capability-named ports, interface narrowing, pass-through layer removal, import-direction repair, DTO or ORM boundary repair, exception translation, transaction-owner clarification, boilerplate deletion, behavior-focused tests, route metadata, and directly synchronized docs or templates | visual-duplication DRY, different domain meanings merged, one-caller interface theater, vague Service, Manager, Helper, Util, Common, Base, or Abstract names, future-only interfaces, fake DDD for CRUD, mode flag helper, excessive DTO conversion, provider error erasure, DI or config larger than logic, hidden proxy or event-bus work, scattered business rule, external SDK or DB row leakage, full-server policy tests, raw payloads deep in app code, provider errors leaking outward, hardcoded policy values, dependency direction inversion, domain importing infrastructure, application importing concrete infrastructure, presentation calling repositories or gateways directly, public contract drift, transaction owner drift, or mocks on private choreography | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Abstraction boundary reviewed, change reason and future scenarios, same-concept gate, candidates compared, public promise and owners, layer dependency contract, layer-integrity report, type-existence and deletion-pass result, over- and under-abstraction findings, test shape, verification, and remaining reversibility risk |
 | Code review or implementation needs module-boundary triage for change spread, change axes, stability direction, co-change clusters, monorepo package direction, package roles, workspace tags, TypeScript path alias boundaries, compile-success versus allowed-dependency confusion, data ownership, policy ownership, failure ownership, import direction, cross-package deep imports, package exports, circular dependencies, DTO leakage, shared/common/utils growth, mock-heavy tests, repeated policy conditions, enum interpretation, repository business logic, anemic domain, domain-to-I/O leakage, transaction boundary mismatch, technical event names, public module API bloat, caller sequencing, premature common helpers, bug/fix distance, config ownership, log responsibility, exception translation, cache invalidation ownership, repeated authorization checks, frontend/backend policy leakage, time policy, batch or worker bypass, shallow modules, or temporary-code accumulation | `.mustflow/skills/module-boundary-review/SKILL.md` | Change reason, change-axis ledger, changed-file spread, co-change evidence, module and package graph evidence, package role and platform tags, `tsconfig.paths`, bundler and test resolver aliases, affected build/test graph evidence, deploy or prune artifact expectations, stability evidence, change-simulation evidence, ownership evidence, test evidence, and configured command intents | Policy ownership, DTO boundaries, mapper boundaries, module public APIs, package `exports`, import direction, workspace dependency direction, package tag constraints, path-alias reach-through repair, config injection, exception translation, cache invalidation ownership, authorization checks, event facts, worker entrypoints, stability-direction repair, shallow-wrapper removal, focused tests, and directly synchronized docs or templates | layer-name theater, role-sliced files that always change together, technical layer split mistaken for change unit, TypeScript path alias mistaken for an architecture boundary, compiling import treated as an authorized dependency, relative sibling package reach-through, app package imported by reusable code, stable package importing volatile app, route, database, provider, or UI details, feature-to-feature internal import, UI kit knowing business state, package deep import, broad export map, root barrel cycle, circular dependency, DTO infection, ownerless shared helper, broad `types` bucket, broad noun module, thin forwarding files, public API wider than hidden behavior, mock-heavy rule tests, copied policy condition, enum rule scatter, repository-owned business rule, service-only domain, domain I/O coupling, cross-owner transaction, table-change event, exposed internals, caller order dependency, unsafe reuse, repeated temporary branch, bug/fix distance, config chaos, misleading logs, leaked provider or DB errors, stale cache owner drift, copied auth checks, frontend reconstructing backend policy, inconsistent time rules, worker bypass, or random capability loss when a module is removed | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Module boundary reviewed, change-spread and co-change evidence, monorepo package graph evidence, resolver graph evidence, change simulations, stability direction, owner and import findings, shallow module findings, boundary fixes or recommendation, evidence level, verification, and remaining module-boundary risk |
 | Code review or implementation needs change-blast-radius triage for maintainability risk from unpredictable next-change spread, historical co-change spread, one change reason scattered across files, multiple change reasons in one file, controller workflow leakage, junk-drawer service names, boolean mode flags, trash-can option objects, scattered domain rules, scattered authorization, hidden state transitions, direct time or randomness, unclear transaction boundaries, external API and DB coupling, retry without idempotency, cache-as-truth decisions, config flag combinations, tenant or partner hardcoding, legacy branches in the core path, DTO/entity/view model mixing, ambiguous nullable values, swallowed exceptions, low-context logs, implementation-coupled tests, mock-heavy tests, decorative abstractions, premature DRY, hidden ordering dependency, invisible event contracts, migration/runtime compatibility, or hard-to-delete features | `.mustflow/skills/change-blast-radius-review/SKILL.md` | User goal, current diff or target files, change-reason ledger, historical co-change ledger, blast-radius ledger, ownership ledger, deleteability ledger, test and operations evidence, and configured command intents | Policy ownership, workflow boundaries, explicit modes, option-object narrowing, authorization owner, state-transition owner, transaction and retry/idempotency boundaries, cache truth boundary, config and tenant variation isolation, legacy adapters, DTO mapping ownership, result types, traceable logs, behavior-focused tests, event contracts, migration compatibility notes, and directly synchronized docs or templates | clean-code theater, unpredictable edit spread, repeated cross-repo co-change, unrelated reasons in one file, controller as boss, junk-drawer service, boolean maze, option combinatorics, copied policy, copied auth, scattered status writes, hidden time or random, partial data after failure, duplicate side effect, stale cache authority, untested flag product, hardcoded customer branch, legacy core pollution, object-language mixing, nullable ambiguity, false success, useless logs, brittle process tests, five-mock class, decorative interface, wrong DRY, order-sensitive line shuffle, ghost event coupling, deploy gamble, or feature that cannot be deleted cleanly | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Change blast radius reviewed, next likely change and owner, historical co-change evidence, spread and deletion path, maintainability findings, fixes or recommendation, verification, and remaining change-spread risk |
 | Code review or implementation needs business-rule-leakage triage for money, permission, ownership, state, settlement, discount, coupon, refund, inventory, notification, subscription, visibility, eligibility, expiry, price, tax, fee, points, reports, tenant scope, UI-only guards, controller eligibility checks, direct status assignment, query predicates as policy, list/detail scope mismatch, admin path bypass, batch hidden policy, tests at the wrong layer, duplicated business constants, date or timezone policy drift, ambiguous `isActive` or `canUse`, authentication/authorization/eligibility mixing, ownership boundary drift, broad update DTOs, PATCH null semantics, mapper logic, default-value drift, misleading error messages, swallowed business failures, transaction/action mismatch, event timing, duplicate requests, webhook trust, out-of-order events, cache-as-rule, search index drift, report or settlement SQL, public text drift, or other bypass entrances | `.mustflow/skills/business-rule-leakage-review/SKILL.md` | User goal, current diff or target files, rule ledger, entrypoint ledger, enforcement ledger, consistency ledger, state/transaction/event/idempotency/default/error evidence, and configured command intents | Shared policy/domain/application/database rule owner, server-side enforcement, reusable query scopes, list/detail consistency, update field restrictions, PATCH intent models, mechanical mappers, default ownership, visible failures, transaction and event boundaries, idempotency, webhook verification, search/detail rechecks, report calculation owners, focused tests, and directly synchronized docs or templates | clean code with leaked money rule, client-only rule, controller judge, scattered status assignment, missing query predicate, URL detail bypass, admin bypass, batch bypass, misplaced test, magic policy number, timezone cutoff drift, vague helper meaning, auth/eligibility blur, user-vs-tenant ownership hole, mass update, PATCH null corruption, mapper policy, default mismatch, error-text lie, false success, half-committed business action, pre-commit event, duplicate refund or coupon use, trusted webhook, stale event transition, wrong cache viewer, dead search result, settlement query drift, support macro drift, or uninspected entrypoint | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Business rule leakage reviewed, source of truth and entrypoints, enforcement and bypass findings, consistency notes, fixes or recommendation, verification, and remaining rule-leakage risk |
 | Payment, checkout, authorization, capture, refund, partial refund, subscription, invoice, credit note, receipt, tax document, trial, grace period, coupon, promotion, inventory reservation, fulfillment, entitlement, settlement, fee, payout, chargeback, dispute, fraud review, card testing, provider webhook, payment session, payment link, payment-provider integration, admin manual payment change, payment log, PCI-sensitive data handling, or payment-related test needs payment-integrity triage for duplicate, late, out-of-order, wrong-actor, wrong-amount, wrong-currency, timeout, retry, idempotency, ledger, reconciliation, or audit risk | `.mustflow/skills/payment-integrity-review/SKILL.md` | User goal, current diff or target files, money-event ledger, provider interaction ledger, state-transition ledger, idempotency and uniqueness ledger, amount and currency ledger, invoice/receipt/tax ledger, dispute and fraud ledger, ownership ledger, fulfillment and entitlement ledger, webhook and retry ledger, audit and sensitive-data ledger, existing tests, and configured command intents | Payment state machines, server-side amount calculation, invoice and line-item snapshots, minor-unit money handling, tax snapshots and reversals, receipt delivery tracking, object ownership checks, idempotency keys, provider ID uniqueness, webhook raw-body signature verification, webhook event dedupe, queue handoff, one-time fulfillment, async payment handling, authorization/capture distinctions, refund/dispute/subscription transitions, inventory and coupon reservation, timeout and retry classification, payout reconciliation, append-only ledgers, secret and payment-data redaction, fraud and card-testing defenses, admin audit trails, stale payment endpoint cleanup notes, focused nightmare-path tests, and directly synchronized docs or templates | paid-boolean shortcut, client-trusted amount, wrong-owner order/payment/refund/subscription ID, amount drift, float money math, missing idempotency, per-retry UUID idempotency, missing provider uniqueness, duplicate webhook, out-of-order webhook, JSON-parsed webhook signature breakage, disabled webhook signature, success-page-as-proof, double fulfillment, async payment premature fulfillment, authorized-treated-as-captured, double refund, missing refund-dispute collision policy, missing tax reversal, receipt delivery gap, chargeback evidence gap, subscription period-only active check, inventory oversell, coupon double spend or lost coupon, timeout-treated-as-failure, blind retry, mutable ledger, payout-as-revenue confusion, card testing exposure, card data logging, test/live secret mix, unaudited admin override, stale payment API, or happy-path-only payment tests | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Payment integrity reviewed, money-event/provider/state/idempotency/amount/invoice/receipt/tax/ownership/fulfillment/webhook/dispute/fraud/payout/audit map, findings, fixes or recommendation, nightmare-path evidence, verification, and remaining payment-integrity risk |
-| Credit, point, wallet balance, reward point, prepaid credit, usage credit, bonus credit, loyalty point, stored-value balance, balance deduction, accrual, refund, reversal, expiration, reservation, capture, release, admin adjustment, transfer, ledger table, balance cache, reconciliation job, settlement report, or credit-related test needs credit-ledger-integrity triage for ledger identity, idempotency, atomic balance changes, concurrency, ordering, ownership, amount precision, rounding, policy snapshots, expiry lots, reservation state, failure recovery, audit evidence, or reconciliation risk | `.mustflow/skills/credit-ledger-integrity-review/SKILL.md` | User goal, current diff or target files, balance surface ledger, ledger-entry ledger, source identity ledger, atomicity ledger, amount and unit ledger, ownership ledger, expiry and lot ledger, reservation ledger, queue and cache ledger, audit and reconciliation ledger, existing tests, and configured command intents | Ledger-entry models, source identifiers, idempotency comparison, conditional balance updates, database constraints, transaction boundaries, row-lock targets, optimistic-lock retry classification, amount validation, rounding policy, non-negative invariants, refund and reversal modeling, partial-use handling, expiry lot allocation, reservation/capture/release transitions, queue idempotency and ordering, cache invalidation, replica-read routing, admin adjustment audit trails, reconciliation checks, evidence logs, focused concurrency and failure tests, and directly synchronized docs or templates | mutable-balance-only path, missing source key, weak idempotency comparison, read-then-update subtraction, unchecked affected rows, split transaction, wrong lock target, optimistic retry double-spend, float credit math, hidden rounding rule, negative or zero amount abuse, app-only non-negative promise, missing unique ledger key, generic balance increase refund, partial refund blind spot, expiry lot loss, expiry batch race, missing reservation state, arbitrary status update, queue reorder damage, duplicate consumer delivery, cache-trusted deduction, replica stale balance, unaudited admin adjustment, request-body wallet ownership, current-price recalculation, missing policy snapshot, no failure injection, no balance-vs-ledger reconciliation, vague deduction log, or happy-path-only credit tests | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Credit ledger integrity reviewed, balance/ledger/source/atomicity/amount/ownership/expiry/reservation/queue/cache/audit/reconciliation map, findings, fixes or recommendation, nightmare-path evidence, verification, and remaining credit-ledger integrity risk |
+| Credit, point, wallet balance, reward point, prepaid credit, usage credit, bonus credit, loyalty point, stored-value balance, purchased versus subscription or promotional lot, balance deduction, accrual, refund, reversal, expiration, rollover, user-visible price quote, reservation, capture, release, admin adjustment, transfer, ledger table, balance cache, reconciliation job, settlement report, or credit-related test needs credit-ledger-integrity triage for ledger identity, idempotency, atomic balance changes, concurrency, ordering, ownership, amount precision, rounding, policy snapshots, balance rights, expiry lots, quote binding, reservation state, failure recovery, audit evidence, or reconciliation risk | `.mustflow/skills/credit-ledger-integrity-review/SKILL.md` | User goal, current diff or target files, balance surface ledger, ledger-entry ledger, source identity ledger, atomicity ledger, amount and unit ledger, ownership ledger, expiry and lot ledger, balance-rights ledger, reservation ledger, quote and fulfillment ledger, queue and cache ledger, audit and reconciliation ledger, existing tests, and configured command intents | Ledger-entry models, source identifiers, idempotency comparison, conditional balance updates, database constraints, transaction boundaries, row-lock targets, optimistic-lock retry classification, amount validation, rounding policy, non-negative invariants, refund and reversal modeling, partial-use handling, balance-class separation, expiry lot allocation, quote binding, maximum reservation, usable-result capture, release, reservation transitions, queue idempotency and ordering, cache invalidation, replica-read routing, admin adjustment audit trails, reconciliation checks, evidence logs, focused concurrency and failure tests, and directly synchronized docs or templates | mutable-balance-only path, missing source key, weak idempotency comparison, read-then-update subtraction, unchecked affected rows, split transaction, wrong lock target, optimistic retry double-spend, float credit math, hidden rounding rule, negative or zero amount abuse, app-only non-negative promise, missing unique ledger key, generic balance increase refund, partial refund blind spot, commingled purchased and expiring rights, expiry lot loss, expiry batch race, stale or unbound quote, over-reservation, unusable-result capture, missing release, missing reservation state, arbitrary status update, queue reorder damage, duplicate consumer delivery, cache-trusted deduction, replica stale balance, unaudited admin adjustment, request-body wallet ownership, current-price recalculation, missing policy snapshot, no failure injection, no balance-vs-ledger reconciliation, vague deduction log, or happy-path-only credit tests | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Credit ledger integrity reviewed, balance/ledger/source/atomicity/amount/ownership/balance-rights/expiry/quote/reservation/queue/cache/audit/reconciliation map, findings, fixes or recommendation, nightmare-path evidence, verification, and remaining credit-ledger integrity risk |
 | Notification generation, email, push, SMS, in-app inbox, digest, reminder, campaign, announcement, marketing message, transactional message, security alert, receipt, legal notice, notification preference, unsubscribe, suppression, quiet hours, timezone schedule, provider webhook, delivery attempt, notification template, or notification audit work needs notification-delivery triage for duplicate, late, suppressed, unsubscribed, wrong-channel, quiet-hours, timezone, retry, provider-webhook, hard-bounce, complaint, invalid-token, fallback, or audit risk | `.mustflow/skills/notification-delivery-integrity-review/SKILL.md` | User goal, current diff or target files, notification event ledger, notification intent ledger, recipient/channel/category ledger, preference and legal policy ledger, suppression ledger, schedule/timezone/quiet-hours/digest ledger, delivery job and attempt ledger, provider event ledger, in-app inbox ledger, audit/security/privacy/operations ledger, existing tests, and configured command intents | Source event and outbox records, notification intent records, preference snapshots, final pre-send rechecks, suppression records, schedule records, timezone-safe recurring intent, quiet-hours behavior, digest records, delivery jobs, delivery attempts, provider event receipts, provider webhook signature and dedupe handling, in-app inbox state, semantic dedupe keys, queue priority, retry classification, template snapshots, redacted audit logs, campaign dry run/sample/canary/ramp-up/kill-switch controls, focused hostile-path tests, and directly synchronized docs or templates | sendNotification shortcut, global opt-out, marketing consent bypass, provider acceptance treated as delivery, missing outbox, stale recipient scope, hard bounce ignored, complaint ignored, one-click unsubscribe missing, link-scanner mutation, stale push token, logout token leak, collapse key misuse, lockscreen data leak, in-app inbox count drift, mark_all_read_before race, semantic dedupe key missing, aggregation hidden as dedupe, digest window drift, quiet-hours delayed blast, timezone DST bug, provider limit starvation, unknown provider outcome blind retry, poison notification loop, provider webhook signature or dedupe gap, template render-time leak, account deletion pending-send leak, fallback spam, dry-run/sample/canary gap, no why-sent or why-suppressed audit, or happy-path-only notification tests | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Notification delivery reviewed, source-event/intent/recipient/preference/suppression/schedule/digest/delivery/provider/inbox/audit map, findings, fixes or recommendation, hostile-path evidence, verification, and remaining notification-delivery risk |
 | Code review or implementation needs api-misuse-resistance triage for APIs, SDKs, function boundaries, service methods, endpoints, command methods, DTOs, request shapes, response shapes, operation names, call ordering, lifecycle setup, boolean parameters, option bags, null or empty semantics, internal table leakage, string-only errors, success-only failure models, idempotency, pagination, sorting, filtering, authorization shape, status mutation, PATCH command buses, time formats, money amounts, open enums, async jobs, bulk partial failures, cacheability, response size, overfragmented calls, internal/external API mixing, version policy, deprecation telemetry, rate limits, retry hints, observability, SDK ergonomics, or caller contract tests | `.mustflow/skills/api-misuse-resistance-review/SKILL.md` | User goal, current diff or target files, caller ledger, operation ledger, shape ledger, compatibility evidence, existing style, and configured command intents | Operation-centered names, explicit lifecycle or builder boundaries, named options or split operations, narrowed option bags, explicit absence and PATCH semantics, public DTO mapping, stable errors, idempotency metadata, cursor stability, auth-specific operations, command-shaped state changes, time and money units, unknown enum handling, job resources, item-level bulk results, cache and rate-limit hints, observability fields, SDK examples, focused tests, and directly synchronized docs or templates | implementation-leaking name, hidden state machine, boolean riddle, trash-can options, null folklore, DB schema frozen as API, string-only error, success-only design, duplicate side effect, moving-list pagination, random default order, hidden permission mode, illegal status transition, PATCH-as-command bus, timezone ambiguity, floating money, closed external enum, fake synchronous completion, erased partial failure, uncacheable expensive read, all-in-one payload bloat, frontend-as-backend call graph, internal/external contract blur, decorative version number, unmeasured deprecation, retry-hostile rate limit, untraceable operation, awkward SDK, happy-path-only contract test, or first-time caller trap | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | API misuse resistance reviewed, caller and operation ledgers, shape and contract findings, fixes or recommendation, compatibility notes, verification, and remaining caller-misuse risk |
 | Code review or implementation needs error-message-integrity triage for error text, error codes, validation messages, parse failures, API or CLI error envelopes, public user messages, internal logs, structured diagnostics, exception wrapping, provider errors, retryable flags, idempotency errors, queue or batch failures, partial failures, permission errors, conflict errors, impossible-state errors, support IDs, redaction, stable machine fields, monitoring fields, or troubleshooting text | `.mustflow/skills/error-message-integrity-review/SKILL.md` | User goal, current diff or target files, error audience ledger, error contract ledger, disclosure ledger, recovery ledger, and configured command intents | Stable error codes, expected/actual fields, failed-operation context, reasons, safe identifiers, public/internal message split, redaction, retryability, idempotency metadata, provider metadata, parse positions, range bounds, conflict facts, partial-failure summaries, structured log fields, focused tests, and directly synchronized docs or templates | empty failed label, invalid-value fog, missing action, result repeated as cause, no work context, public/internal message mixing, sensitive value leak, missing safe identifier, retry ambiguity, try-again-later dodge, unstable string-only code, overbroad error bucket, validation info leak, parse location missing, range without bounds, timezone ambiguity, provider evidence loss, cause destruction, brittle message concatenation, prose-only logs, internal jargon in user text, permission existence leak, should-never-happen message, vague conflict, idempotency uncertainty, missing attempt count, partial failure erased, untested error contract, no 30-second next action, or call-site-specific taxonomy drift | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `test_audit`, `docs_validate_fast`, `test_release`, `mustflow_check` | Error message integrity reviewed, surfaces and audiences, code/message contract, evidence and redaction findings, fixes or recommendation, verification, and remaining error-message risk |
@@ -615,7 +708,7 @@ routes. Event routes stay inactive until their event occurs.
 | Code review, implementation, runbook work, or release preparation needs deployment-rollout safety review for server, backend, worker, scheduler, queue, cron, container, VM, serverless, DB migration, config, feature flag, cache, deployment pipeline, release envelope, image digest, deployment history, traffic rollback, canary, rollback, health check, readiness/liveness/startup probe, graceful shutdown, artifact promotion, release observability, or post-deploy smoke behavior where the deploy must be rolled out, stopped, observed, and rolled back safely | `.mustflow/skills/deployment-rollout-safety-review/SKILL.md` | Deployment resource ledger, release envelope, artifact identity, environment promotion path, deployment model, compatibility matrix, config diff, migration order, rollback history, traffic rollback path, cache and message compatibility, probe model, shutdown and drain behavior, canary cohort, version-split telemetry, stop conditions, rollback limits, synthetic transactions, post-deploy metrics, and configured command intents | Runbooks, release checklists, pipeline metadata, smoke tests, probe tests, config validation, feature-flag defaults, cache-key versions, worker-drain handling, deployment attribution, rollback compatibility notes, focused tests, and directly synchronized templates | unknown blast radius, missing release id, mutable latest tag, tag without digest, per-environment rebuild drift, deleted rollout history, cold old version, traffic rollback tied to rebuild, code and migration lockstep, destructive rollback SQL overclaim, missing PITR practice, config in-place mutation, missing startup config validation, process-only health check, readiness/liveness/startup probe collapse, liveness restart loop, ungraceful shutdown, load balancer drain shorter than app shutdown, worker work loss, non-idempotent queue retry, N-1 message incompatibility, unknown event poison message, missing external compensation, API N-1 or N+1 break, missing kill switch, unsafe flag fallback, vague canary cohort, global-average canary metrics, no automatic stop condition, read-only smoke, log format alert breakage, blanket cache flush, scheduler duplicate execution, CRD or operator downgrade break, missing deployment lock, production command without dry-run, or code-only rollback overclaim | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Deployment rollout boundary reviewed, resource ledger and release envelope, artifact identity, config/migration/cache/queue/API/probe/shutdown/canary/rollback/observation findings, verification, and remaining deployment-rollout risk |
 | Code review, implementation, runbook work, or infrastructure review needs cloud-cost-guardrail review for cloud accounts, projects, subscriptions, environments, Kubernetes namespaces, serverless, databases, object storage, block storage, snapshots, NAT, private endpoints, public IPs, egress, CDN, logs, metrics, traces, autoscaling, quotas, budgets, tags, temporary resources, container registries, Marketplace, LLM APIs, external APIs, or third-party SaaS where spend must be attributed, capped, lifecycle-managed, alerted, and safely stoppable before a silent bill explosion | `.mustflow/skills/cloud-cost-guardrail-review/SKILL.md` | Cost surface ledger, budget actual and forecast thresholds, automated non-production action path, account or project isolation, quota and cap model, tag taxonomy, temporary resource expiration, network cost model, telemetry cost model, storage lifecycle model, commitment baseline, Marketplace or LLM usage limits, and configured command intents | Cost guardrail docs, infrastructure policy files, review checklists, tag schemas, quota notes, budget-action runbooks, cleanup rules, retention defaults, autoscale caps, Kubernetes ResourceQuota and LimitRange notes, registry lifecycle policies, provider usage caps, focused tests, and directly synchronized templates. CI runner minutes, workflow matrix cost, artifact retention, cache quota, and release asset handoff route to `ci-pipeline-triage` first | notification-only budget, imagined hard spending limit, mixed prod and dev account, over-wide service quota, missing owner tag, tag-key chaos, no expires_at, stopped VM with NAT or DB still running, unbounded autoscale, missing Kubernetes ResourceQuota, inflated requests growing nodes, cloud-native service through NAT, untracked egress, cross-AZ surprise, idle public IPv4, no CDN cache cost control, log ingest flood, infinite retention, high-cardinality metric label, unbounded flow or audit logs, object lifecycle missing, cold-storage minimum-duration trap, stale block volume type, snapshot landfill, sticky DB storage growth, unbounded registry images, premature commitment, stateful spot misuse, unmonitored Marketplace or LLM spend, CI billing routed to broad cloud review before localizing workflow cost, or no safe cost stop runbook | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Cloud cost boundary reviewed, cost surface ledger, budget and action model, isolation/quota/tag/autoscale/Kubernetes/network/telemetry/storage/registry/commitment/spot/Marketplace/LLM/SaaS guardrail findings, manual-only provider checks, verification, and remaining cloud-cost risk |
 | Code review or implementation needs rate-limit integrity triage for rate limits, throttling, quotas, API usage limits, request costs, token bucket, leaky bucket, fixed window, sliding window counter, sliding window log, GCRA, Redis counters, Lua or EVAL updates, CDN or WAF limits, gateway limits, service limits, tenant, user, API key, route group, IP, 429, `Retry-After`, `RateLimit`, shadow mode, operator reset, async enqueue, cached-hit counting, or concurrency-limit overlap that must protect a named resource without bypass, unfairness, counter drift, storage growth, retry storms, or misleading client hints | `.mustflow/skills/rate-limit-integrity-review/SKILL.md` | Protected resource ledger, cost-weighted request ledger, layer model, key model, algorithm and storage model, failure mode model, response contract, observability and operator evidence, and configured command intents | Protected-resource definitions, request cost weights, per-key policy, layered limit placement, route-template keys, atomic counter updates, TTLs, storage-time use, fail-open or fail-closed policy, blocked-decision cache, shadow mode, 429 response shape, observability fields, operator lookup or reset behavior, focused tests, and directly synchronized docs or templates | algorithm-first limiter, request-count-only quota, IP-only authenticated key, raw URL key explosion, missing identity-header policy, fixed-window boundary burst, costly sliding-window log on hot paths, non-atomic Redis read-modify-write, missing counter TTL, Redis Cluster hash-slot failure, app-server clock reset drift, process-local global quota, approximate edge limit treated as precise, hidden fail-open, free failed responses, rate versus concurrency confusion, unhelpful or leaky 429, synchronized retry wave, unsafe allow-decision cache, no shadow-mode ramp, missing policy id logs, raw Redis reset, unlimited async enqueue, cached CDN hit ambiguity, or rate limit treated as authorization or hard cost control | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Rate-limit policy boundary reviewed, protected resource and cost/layer/key/storage/fail-mode/response/operator model mapped, evidence level, verification, and remaining rate-limit-integrity risk |
-| Code review or implementation needs idempotency-integrity triage for repeated or out-of-order requests, retries, webhooks, queue delivery, scheduler or batch reruns, callbacks, timeout recovery, duplicate business commands, stale asynchronous completions, aggregate sequence gaps, or old decisions that can repeat effects or overwrite newer authority | `.mustflow/skills/idempotency-integrity-review/SKILL.md` | Operation, attempt, event, message, effect, actor, tenant, target, and provider identity ledger; canonical payload binding; ordering aggregate and scope; state, version, sequence, generation, and fence authority; durable dedupe; response replay; recovery; queue, webhook, scheduler, and batch evidence; test evidence; and configured command intents | Stable cross-retry operation identity, distinct attempt and effect IDs, canonical payload conflict checks, unique non-null durable admission, atomic insert-or-return, aggregate versions, bounded gap handling, conditional state or fence writes, affected-row decisions, inbox and outbox records, result lookup and replay, processing recovery, invariant reconciliation, adversarial duplicate and stale-order tests, and directly synchronized docs or templates | per-retry operation ID, nullable uniqueness key, key not bound to payload or actor, memory-only or TTL-only dedupe, app-only check-then-insert, global-order assumption, timestamp or serial ID as causality, unowned sequence gap, unconditional late write, blind CAS retry, cancellation as rollback, single-flight as correctness, lock without fencing, ambiguous effect replay, queue settlement split, duplicate compensation, or ordinary tests presented as concurrency proof | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Idempotency boundary reviewed, logical operation and attempts, duplicate sources, ordering aggregate and authority token, gap and stale-result policy, durable dedupe and replay, external-effect and reconciliation findings, adversarial evidence, verification, and remaining idempotency-integrity risk |
+| Code review or implementation needs idempotency-integrity triage for repeated or out-of-order requests, retries, replans, webhooks, queue delivery, scheduler or batch reruns, callbacks, timeout recovery, duplicate business commands, stale asynchronous completions, aggregate sequence gaps, or old decisions that can repeat effects or overwrite newer authority | `.mustflow/skills/idempotency-integrity-review/SKILL.md` | Workflow, plan, step, operation, attempt, event, message, effect, actor, tenant, target, and provider identity ledger; canonical payload binding; ordering aggregate and scope; state, version, sequence, generation, and fence authority; durable dedupe and retention lifecycle; response replay; UNKNOWN and RECONCILING recovery; queue, webhook, scheduler, and batch evidence; test evidence; and configured command intents | Stable cross-retry and cross-replan operation and effect identity, distinct attempts, canonical payload conflict checks, opaque trusted-boundary keys, unique non-null durable admission, lifecycle-aligned retention, atomic insert-or-return, aggregate versions, bounded gap handling, conditional state or fence writes, inbox and outbox records, result lookup and replay, processing recovery, invariant reconciliation, adversarial duplicate and stale-order tests, and synchronized docs or templates | per-retry or per-replan effect ID, nullable uniqueness key, personal data in idempotency key, key not bound to payload or actor, memory-only or arbitrary-TTL dedupe, record expiry during reconciliation, app-only check-then-insert, timestamp causality, unowned sequence gap, stale write, blind UNKNOWN retry, single-flight as correctness, lock without fencing, ambiguous effect replay, queue split, duplicate compensation, or ordinary tests presented as concurrency proof | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Idempotency boundary reviewed, workflow/plan/step/effect/attempt identities, duplicate sources, retention, ordering and authority, durable dedupe and replay, UNKNOWN reconciliation, adversarial evidence, verification, and remaining idempotency-integrity risk |
 | Code review or implementation needs retry-policy integrity triage for retry loops, SDK retry configs, client middleware, `while true`, `for (;;)`, recursive retry, `maxAttempts`, `maxRetries`, `maxElapsedTime`, deadline, timeout, sleep, backoff, jitter, `Retry-After`, retry predicates, layered retries, circuit breakers, bulkheads, token buckets, queue redelivery, broker retry, cancellation-aware sleep, or retry observability that can amplify failures, duplicate side effects, hide permanent errors, exhaust pools, or overload dependencies | `.mustflow/skills/retry-policy-integrity-review/SKILL.md` | Retry surface, layered retry ledger, attempt budget, retry predicate, side-effect and idempotency ledger, backoff and jitter policy, overload and throttling evidence, observability and test evidence, and configured command intents | Bounded attempts, max elapsed time, per-attempt timeout, total deadline, cancellation propagation, retry predicates, exponential backoff with jitter, `Retry-After` parsing and clamping, idempotency key reuse, dependency-specific policy, retry wrapper diagnostics, per-attempt logs and metrics, focused retry tests, and directly synchronized docs or templates | retry amplification, infinite retry, capped backoff without stop condition, timeout gap for DNS, TLS, pool wait, streaming, or parsing, fixed-sleep herd behavior, broad catch-and-retry, permanent error retry, unknown-outcome replay, new idempotency key per attempt, key not bound to actor or payload, retry inside transaction or lock, pool starvation, unlimited parallel retry, stale per-key failure counter, global limiter unfairness, wrong circuit breaker or bulkhead order, wrapper losing cause/status/retry-after/request id, committed-response retry, non-replayable streaming body retry, app-plus-broker retry multiplication, cancellation-ignoring sleep, generic dependency policy, missing retry metrics, or happy-path-only retry tests | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Retry policy boundary reviewed, layer multiplication and attempt budget mapped, timeout/backoff/predicate/idempotency/throttling findings, evidence level, verification, and remaining retry-policy-integrity risk |
 | Code review or implementation needs queue-processing integrity triage for queues, streams, pub/sub handlers, workers, task runners, consumers, producers, webhook handoffs, DLQ replayers, retry workers, ack, nack, reject, delete, visibility timeout, offset commit, publisher confirm, prefetch, batch commit, rebalance, FIFO message group, deduplication, or worker-loss behavior that can lose messages, duplicate side effects, hide poison messages, reorder state, exhaust consumers, or falsely claim processing success | `.mustflow/skills/queue-processing-integrity-review/SKILL.md` | Broker and delivery model, success boundary, producer boundary, consumer state ledger, failure and retry policy, concurrency and ordering evidence, observability evidence, test evidence, and configured command intents | Settlement timing, publisher confirmation, outbox or inbox record, durable message key, conditional state transition, bounded retry and DLQ policy, visibility extension, prefetch and concurrency bounds, rebalance handling, shutdown drain, focused queue replay tests, and directly synchronized docs or templates | ack-before-work, catch-and-ack, finally-ack, auto-ack, stale receipt handle, premature offset commit, batch partial-failure skip, unbounded requeue, poison-message loop, unsafe visibility timeout, in-flight saturation, FIFO group misuse, worker-loss false success, producer publish split, side-effect duplication, missing rebalance fence, unlimited parallelism, DLQ bucket without replay policy, or missing decision-point observability | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Queue processing boundary reviewed, broker model and settlement evidence, producer/consumer/retry/DLQ/order/concurrency findings, evidence level, verification, and remaining queue-processing-integrity risk |
 | Code review or implementation needs transaction-boundary integrity triage for transactions, ORM atomic blocks, unit-of-work code, database write paths, service workflows, command handlers, webhook processors, queue consumers, framework transaction annotations, isolation levels, lock usage, rollback behavior, after-commit side effects, outbox patterns, retry handling, or transactional tests that can break a business invariant even when a transaction exists | `.mustflow/skills/transaction-boundary-integrity-review/SKILL.md` | Business invariant, transaction boundary, decision ledger, durable guard evidence, framework behavior, side-effect ledger, failure and retry evidence, test evidence, and configured command intents | Whole read-decision-write transaction boundaries, durable constraints, atomic upsert or conditional update, affected-row checks, version checks, correct lock target, transaction narrowing, after-commit or outbox side effects, idempotent retry classification, focused tests, and directly synchronized docs or templates | app-only `exists()` or `count` guard, stale read-decision-write, absent-row `FOR UPDATE` gap, `SKIP LOCKED` consistency misuse, READ COMMITTED snapshot myth, REPEATABLE READ engine mismatch, SERIALIZABLE without full retry, deadlock or serialization failure treated as plain 500, swallowed rollback trigger, Spring `rollbackFor` miss, self-invocation bypass, `readOnly` write assumption, inner rollback-only surprise, `REQUIRES_NEW` pool pressure, `NESTED` savepoint confusion, Django nested `atomic()` durability myth, pre-commit email/cache/queue side effect, HTTP API inside transaction, Hibernate flush-as-commit confusion, SQLAlchemy implicit transaction surprise, missing optimistic lock, advisory lock scope leak, wrong transaction manager, or transactional test hiding commit behavior | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Transaction boundary reviewed, invariant and decision ledger, durable-guard/lock/isolation/retry/rollback/side-effect findings, evidence level, verification, and remaining transaction-boundary integrity risk |
@@ -702,7 +795,7 @@ routes. Event routes stay inactive until their event occurs.
 | Code, configuration, docs, templates, logs, telemetry, traces, baggage, behavior analytics, credentials, data flows, data residency policy, region or processing-location claims, AI-generated code, authentication, authorization, client-only permission checks, admin operations, audit logs, cache policy, cache-as-authority decisions, claim or policy data, comparison or affiliate data, user-generated content, sessions, tokens, uploads, downloads, signed URLs, API responses, webhooks, job queues, external API call records, external requests, third-party data-use terms, runtime security patch policy, vulnerability scanner advisories, development servers, test UI servers, deployment settings, dependencies, cryptography, secure transport, scanner gates, security invariants, or agent configuration affect secrets, personal data, retention, access control, vendor disclosure, or external disclosure | `.mustflow/skills/security-privacy-review/SKILL.md` | Changed files, sensitive surfaces, actor and resource owner, data-owner boundary, data residency and processing-location boundary, runtime patch boundary, dev-server host or privileged API boundary, AI gateway or budget boundary, server-side authorization rule, file upload/download boundary, API response field boundary, behavior analytics surface, trace or baggage surface, webhook or external-call record surface, admin operation surface, audit-log surface, cache visibility and authority policy, claim or affiliate policy surface, session or token surface, external target, dependency source, advisory exploit preconditions, third-party data-use or terms surface, cryptography or transport surface, scanner evidence, agent-tool permission, deployment setting, project secret and privacy rules, public or packaged surfaces, and command contract entries | Sensitive data handling, authorization, admin operations, data residency, runtime patchability, dev-server file-serving and privileged API gates, AI budget records, behavior analytics, observability identifiers, webhook receipts, external-call records, dead-letter records, audit logs, shared-cache behavior, cache-authority behavior, claim and affiliate disclosure, sessions, tokens, inputs, files, signed URLs, API responses, logs, receipts, generated state, docs, templates, package metadata, deployment settings, and reports | secret leak, personal-data exposure, access-control bypass, client-trusted role or owner value, unsafe admin action, private file exposure, exposed local tool API, over-broad API response, shared-cache leak, unsafe cache authority, unprovable data location, unpatchable runtime, devDependency alert dismissed despite network exposure, privacy-heavy telemetry, unsafe baggage propagation, unsafe webhook payload retention, unsafe external request, supply-chain drift, weak cryptography, insecure transport, protocol parser DoS, over-privileged agent, risky third-party terms, or misleading privacy claim | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | Sensitive surfaces reviewed, data residency, runtime patchability, dev-server/test-UI exposure, AI hard-limit, behavior analytics, observability, and audit boundaries, webhook, external-call, and dead-letter boundaries, cache authority and disclosure boundaries, assumptions checked, disclosure and retention paths, authorization, file, API response, third-party terms, and external-boundary notes, verification, and remaining security or privacy risk |
 | Real or plausible secrets, tokens, credentials, private keys, passwords, session values, service-account values, connection strings, signing secrets, webhook secrets, certificate keys, recovery codes, or production-like credential material appear in files, artifacts, logs, command output, screenshots, fixtures, docs, templates, package output, caches, run receipts, or final reports | `.mustflow/skills/secret-exposure-response/SKILL.md` | Exposure surface, secret type without value, tracked/generated/public/package status, allowed remediation scope, rotation or revocation boundary, and command contract entries | Redaction, omission, placeholder replacement, docs, fixtures, templates, examples, package inputs, generated artifacts, and final report wording | repeated exposure, false fake-value claim, redaction mistaken for revocation, package leak, screenshot leak, history exposure, or secret printed in reports | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | Exposure surfaces reviewed, secret value omitted, remediation made, remaining rotation/revocation/history/external risks, verification, and remaining exposure risk |
 | Security-sensitive behavior changes need abuse-case regression tests | `.mustflow/skills/security-regression-tests/SKILL.md` | Changed boundary, actors, resource ownership, state-changing route, token, file, cryptography, transport, scanner, or invariant behavior, business rule, and expected deny behavior | Test files and related security boundary source | false confidence, happy-path-only coverage, unsafe authorization, token, file, business-rule, cryptography, transport, deployment, or invariant coverage | `test`, `test_related`, `test_audit`, `lint`, `build` | Security boundary, abuse case, defensive test data, tests added or reused, and remaining risks |
-| Outside text, generated content, logs, issues, webpages, pasted prompts, agent rules, MCP/tool configuration, or AI context sources include instructions that could override repository rules, broaden tool access, leak data, or change scope | `.mustflow/skills/external-prompt-injection-defense/SKILL.md` | External text source, direct user request, repository instruction files, conflicting instruction, context sources, tool permission surface, hidden content evidence, and command contract entries | Prompts, fixtures, docs, tests, skills, templates, agent configs, tool configs, and reports that handle untrusted text | prompt injection, context leakage, scope drift, unsafe command authority, or over-broad tool permission | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | External sources reviewed, unsafe instructions neutralized, context and permission boundaries checked, safe requirements adapted, verification, and remaining prompt-injection risk |
+| Outside text, generated content, logs, issues, webpages, pasted prompts, retrieved documents, tool results, agent rules, MCP/tool configuration, or AI context sources can override trusted intent, inject attacker-controlled instructions or security-critical data, broaden capabilities, leak data, or change scope | `.mustflow/skills/external-prompt-injection-defense/SKILL.md` | Authenticated user intent, trusted tool contract, untrusted sources, planner/collector/policy/executor/memory-writer boundaries, typed collector output, field-level provenance and taint, deterministic argument policy, credential ownership, corroboration or approval rule, context and capability surface, hidden content evidence, and configured command intents | Prompts, collectors, typed facts, provenance and taint, deterministic gates, structured executor inputs, server-held credentials, process and capability separation, memory admission, fixtures, docs, tests, skills, templates, agent and tool configs, and reports | instruction injection, data injection, forged metadata or origin, tainted high-impact argument, raw retrieved text reaching executor, collector with write authority, credential exposure, memory poisoning, detector widening capability, context leakage, or scope drift | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | External and runtime sources reviewed, trusted intent separated from untrusted data, typed provenance and taint preserved, capability and memory boundaries checked, deterministic or human gates applied, verification, and remaining prompt/data-injection risk |
 | External code, prose, snippets, scripts, command examples, docs text, prompts, assets, tests, fixtures, schemas, configs, generated patches, or AI-generated material may be copied, adapted, translated, shipped, or preserved in public or packaged repository surfaces | `.mustflow/skills/provenance-license-gate/SKILL.md` | External source, snapshot or revision, destination surface, material type, copy extent, license evidence, attribution requirement, package/public/executable status, and command contract entries | Copied or adapted material, attribution or notices, docs, tests, templates, package metadata, and third-party notice surfaces | unknown-license copy, incompatible license, missing attribution, copied expression mistaken for idea, generated derivative risk, package notice drift, or provenance gap | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | Sources reviewed, copy extent classified, license and attribution decision, adopted/rewritten/omitted material, synchronized notice surfaces, verification, and remaining provenance risk |
 
 ### Data and External Systems
@@ -799,6 +892,23 @@ routes. Event routes stay inactive until their event occurs.
 | Multiple AI workers, subagents, external agents, parallel task runners, or worktree-based worker roles are planned or used for one repository task | `.mustflow/skills/multi-agent-work-coordination/SKILL.md` | Task goal, worker roles, write permissions, file ownership, workspace isolation, credential boundary, merge owner, and command contract entries | Coordination plan, worker instructions, ownership boundaries, merge notes, and directly synchronized tests or docs | same-file races, conflicting instructions, leaked credentials, shared auth cache, untrusted worker output, merge drift, or unverified parallel result | `changes_status`, `changes_diff_summary`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Worker limit, role map, write ownership, isolation and credential boundaries, merge owner, verification, skipped checks, and remaining coordination risk |
 | Brainstorming, option comparison, outside AI advice, planning notes, or loose proposals need evidence-based apply, defer, reject, or research decisions before implementation | `.mustflow/skills/idea-triage/SKILL.md` | User goal, idea list or recommendation, current repository evidence, constraints, and decision mode | Analysis, roadmap entries, and at most one selected follow-up when requested | idea spam, speculative roadmap, current-behavior claims for deferred work, or ungrounded prioritization | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `mustflow_check` | Decision mode, evidence, constraints, option decisions, selected next action, verification needs, and remaining uncertainty |
 | Analysis or a decision record is the current deliverable, the problem has both material uncertainty and material consequences, and no narrower primary skill owns the complete problem | `.mustflow/skills/complex-decision-analysis/SKILL.md` | User request, decision horizon, decision owner, repository evidence, existing skill routes, available evidence sources, and limits on freshness, access, authority, verification, or command execution | Evidence-backed decision records, planning artifacts when requested, and exactly one smallest reversible next action before handoff | universal reasoning skill bloat, analysis paralysis, private scratch reasoning, stale evidence, overconfident causal story, hidden high-sensitivity unknown, or implementation without a narrower handoff skill | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `mustflow_check` | Decision state, problem contract, evidence ledger, baseline, causal model, option comparison, counterargument, decision-reversing evidence, recommendation, smallest reversible next action, handoff skill, verification, and residual risk |
+| A team must decide whether an AI agent, workflow automation, internal tool, or integration is worth building, expanding, replacing, or retiring from expected cost per accepted outcome, human alternatives, supervision, failure recovery, maintenance, break-even volume, throughput value, effective lifetime, NPV, and hard safety requirements | `.mustflow/skills/automation-investment-case-review/SKILL.md` | Accepted-outcome contract, automation cost ledger, human or current-system comparator, quality and failure distribution, volume and demand evidence, fixed and recurring investment, effective lifetime, discount and scenario assumptions, independent safety gate, and configured command intents | Decision records, cost and scenario models, break-even, payback and NPV analysis, sensitivity thresholds, evidence labels, bounded pilot criteria, safety gates, calculation tests, docs, route metadata, and synchronized templates | model-call cost as total cost, attempted task as accepted outcome, perfect-path automation versus fully loaded human comparison, supervision or incident cost omitted, throughput monetized without demand, one pilot generalized, infinite useful life, safety traded for ROI, or universal example numbers embedded as policy | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | Accepted outcome and comparator, cost model, assumptions and evidence levels, break-even/payback/NPV/sensitivity, throughput value, safety-gate result, reversible experiment, verification, and remaining investment uncertainty |
+| Pre-account core-value experiences, account-first versus try-first flows, signup checkpoints, anonymous-to-account transfer, authentication choices, signup or first-run questions, intent personalization, samples, tutorials, contextual help, progressive disclosure, primary calls to action, activation definitions, time-to-value, retained use, paid conversion, or onboarding experiments are created, changed, reviewed, or reported | `.mustflow/skills/product-onboarding-activation-review/SKILL.md` | Eligible visitor or user cohort and assignment, pre-account and account-checkpoint ledger, anonymous transfer, account and identity boundary, question and personalization ledgers, instruction path, sample-to-own contract, user-owned activation postcondition, funnel stages, experiment design, delayed outcomes, guardrails, privacy, and configured command intents | Account checkpoints, guest sandboxes, stable identity and transfer contract, onboarding state and question schemas, authentication choices, intent mappings, editable samples, contextual help or tutorials, focused primary action, assignment and exposure events, activation metrics, fixtures, tests, docs, route metadata, and synchronized templates | signup rate optimized instead of visitor-to-value, sensitive guest operation, anonymous work lost at signup, email used as identity key, unsafe auto-link, authentication UX confused with assurance, universal auth winner, onboarding completer as denominator, tutorial completion counted as activation, profile survey before value, decorative personalization, blank restart after demo, multi-feature choice overload, post-hoc segment win, or borrowed benchmark treated as forecast | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Eligible cohort and activation contract, pre-account and checkpoint decision, transfer and identity boundary, authentication and instruction-path choice, question friction, personalization and sample-to-own decisions, delayed outcome and guardrail evidence, verification, and remaining onboarding-activation risk |
+| Subscription cancellation flows, churn reasons, save offers, downgrade, pause, trial extension, discount, usage credit, dormant-user reactivation, win-back campaigns, offer sequencing, resumed billing, retention, CLTV, or contribution-profit claims are created, changed, reviewed, or reported | `.mustflow/skills/subscription-retention-profit-review/SKILL.md` | Eligible lifecycle and state ledgers, expected usage cadence, churn reason, treatment and cost ledgers, no-offer or business-as-usual holdout, assignment and exposure, natural return, pause and resume contract, contribution-margin horizon, and configured command intents | Retention policy, reason routing, offer eligibility and cooldown, holdout assignment, pause lifecycle, downgrade and credit rules, contribution metrics, event schemas, fixtures, tests, docs, route metadata, and synchronized templates | save rate as profit, acceptor-only denominator, natural return credited to campaign, active payer discounted unnecessarily, fixed last-login threshold, pause counted as active retained, first resumed payment treated as durable recovery, offer parade blocking cancellation, credit cost omitted, discount farming, or observational vendor benchmark embedded as policy | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Lifecycle and eligibility, holdout and causal evidence, reason-matched treatment, incremental contribution components, pause and post-resume survival, cancellation boundary, verification, and remaining subscription-retention profit risk |
+| Weekly or periodic value reports, usage or artifact recaps, saved-time claims, streaks, attendance rewards, grace or recovery, abandoned-work reminders, completion nudges, notification timing, or post-activation engagement experiments are created, changed, reviewed, or reported | `.mustflow/skills/product-engagement-retention-review/SKILL.md` | Eligible cohort or task, assignment and holdout, product cadence, value-report evidence, streak and reward contract, abandoned-work state, contact and cancellation ledger, meaningful downstream outcomes, quality and opt-out guardrails, and configured command intents | Report composition and suppression, value-estimation rules, streak qualification and recovery, reward isolation and withdrawal, task eligibility, reminder timing and cancellation, experiment events, metrics, fixtures, tests, docs, route metadata, and synchronized templates | raw activity presented as value, invented time saved, zero-activity cancellation reminder, attendance-only streak, daily cadence forced on episodic value, reward farming, reward claim treated as habit, natural return credited to reminder, universal ten-minute or one-day timing, stale reminder after completion, multi-send fatigue, return click without completion, unsubscribe or block harm, or observational case study embedded as policy | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Eligible cohort or task and holdout, value-report and estimation contract, streak cadence and quality, reward persistence, abandoned-work timing and cancellation, downstream quality and contact guardrails, verification, and remaining engagement-retention risk |
+| Credit-pack offers, offer timing, price discount versus bonus-credit framing, pack count or spacing, recommendations, purchased-credit expiry, subscription-credit rollover, promotional balances, spend order, credit price disclosure, variable-cost estimates, quote and reservation UX, breakage, repurchase, retention, or credit-monetization experiments are created, changed, reviewed, or reported | `.mustflow/skills/credit-monetization-integrity-review/SKILL.md` | Eligible cohort and pre-trigger assignment, offer and value ledgers, pack and output-equivalent ledger, balance-rights classes, quote and execution contract, short- and long-horizon economics, platform and jurisdiction evidence, experiment design, and configured command intents | Offer timing and suppression, equivalent-value framing, pack architecture and recommendation, balance-class expiry and rollover policy, spend order, exact or bounded price authorization, quote-reserve-settle contract, contribution metrics, fixtures, tests, docs, route metadata, and synchronized templates | post-trigger survivor denominator, signup interruption ignored, fake urgency, unequal discount and bonus called copy test, credits without output meaning, universal three-pack or copied price ratio, dominated decoy, purchased value expired against authority, breakage as profit, balance classes commingled, durable top-up spent before expiring credits, retroactive devaluation, hidden or fake-exact deduction, provider cost charged without usable result, unknown outcome captured, or early purchase rate promoted before contribution and retention | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Eligible cohort and offer timing, economic-equivalence calculation, pack architecture, balance-rights and rollover policy, quote-reserve-settle contract, contribution and trust guardrails, specialist handoffs, verification, and remaining credit-monetization risk |
+| Card-required or cardless trials, paid starters, lifetime or founder access, local-currency presentment, regional purchasing-power pricing, subscription versus usage-based or hybrid monetization, price levels, elasticity, annual prepay, annual discounts, or pricing contribution claims are created, changed, reviewed, or reported | `.mustflow/skills/pricing-model-integrity-review/SKILL.md` | Eligible cohort and pre-gate assignment, value and cost ledger, trial consent and renewal contract, lifetime obligation and heavy-tail ledger, currency and regional eligibility, model and meter contract, price packages, annual-prepay economics, experiment design, and configured command intents | Trial credential policy, consent and reminders, bounded lifetime rights and reserve assumptions, local currency and regional bands, model and value metric, meter and spend controls, price experiments, annual discounts, events, fixtures, tests, docs, route metadata, and synchronized templates | trial-to-paid survivor bias, hidden negative-option renewal, unlimited variable-cost lifetime rights, launch cash treated as profit, PPP treated as willingness to pay, local currency confused with a regional discount, IP-only pricing eligibility, affordability treated as fraud control, internal tokens exposed as consumer value, payer ARPPU replacing eligible contribution, copied price endings, annual cash confused with recognized revenue, or universal discount embedded as policy | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Eligible cohort and trial result, lifetime rights and stress economics, currency and regional policy, monetization model and meter, price elasticity and retained contribution, annual cash and obligation boundaries, specialist handoffs, verification, and remaining pricing-model risk |
+| Paid revives, rewarded-ad revives, lives, energy, natural recovery, credit refills, unlimited play, VIP or subscription benefits, failure monetization, content-consumption pacing, pay-to-win boundaries, IAP cannibalization, ARPDAU, retained play, or player LTV are created, changed, reviewed, or reported | `.mustflow/skills/game-economy-monetization-review/SKILL.md` | Game loop and failure meaning, revive causes and outcomes, energy and content supply, ad and credit rights, membership benefits, payer migration, experiment assignment, economy versions, fairness guardrails, and configured command intents | Recovery and revive policy, energy pacing, rewarded-ad and credit differentiation, membership limits, content-burn metrics, economy events, experiments, fixtures, tests, docs, route metadata, and synchronized templates | ranked outcome sold, meaningful death erased, defect monetized, first fun blocked by depletion, repeat revive spam, ad and credits perfectly interchangeable, forced or unavailable ad blocking play, unlimited membership cannibalizing IAP, finite content exhausted, payer migration hidden, ad-viewer self-selection, or ARPDAU promoted before long-horizon contribution | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Loop and failure boundary, revive and energy policy, ad-credit differentiation, membership and content-life economics, fairness and payer migration, eligible-player contribution, verification, and remaining game-economy risk |
+| Free-tier generosity, hard or soft limits, ad-supported access, interstitial or rewarded-ad placement, first-value ad suppression, frequency caps, result gates, premium ad removal, free-user variable cost, ad cannibalization, conversion, retention, or eligible-user contribution are created, changed, reviewed, or reported | `.mustflow/skills/freemium-ad-monetization-review/SKILL.md` | First owned value, free-tier and variable-cost ledger, ad format and placement, natural transition, load and failure state, frequency and suppression, paid entitlement, privacy and platform evidence, experiment assignment, and configured command intents | Free-value and repeat-limit policy, interstitial and rewarded-ad placement, first-value suppression, failure-safe continuation, frequency caps, premium treatment, contribution metrics, fixtures, tests, docs, route metadata, and synchronized templates | free amount and ads collapsed into one decision, first value blocked, active work interrupted, completed result held hostage, forced ad called rewarded, ad load blocking progress, processing delay manufactured, copied frequency default, result watermark or owned cross-promotion mistaken for ad inventory, gross ad revenue treated as profit, payer displacement hidden, or SDK privacy and performance cost omitted | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Eligible cohort and first value, free allowance and cost, ad format placement frequency and failure path, distribution-policy handoff, conversion and cannibalization, privacy and performance guardrails, verification, and remaining freemium-ad risk |
+| Inviter-only, invitee-only, dual-sided, tiered, or milestone customer referral rewards, attribution, invite codes or links, valid-referral qualification, pending or vested rewards, reversal, self-referral, reward farming, rolling thresholds, incrementality, or referred-user contribution are created, changed, reviewed, or reported | `.mustflow/skills/referral-incentive-integrity-review/SKILL.md` | Inviter and invitee eligibility, attribution and conflict ledger, downstream qualification, reward states and rights, tier window and caps, multi-signal abuse evidence, legitimate collisions and appeals, causal experiment design, and configured command intents | Customer referral eligibility and attribution, qualification, pending vesting grant and reversal, tiering, caps, reason codes, anti-abuse and appeal policy, contribution metrics, fixtures, tests, docs, route metadata, and synchronized templates | customer referral confused with affiliate or partner commission, reward direction confused with tiering, signup treated as valid referral, unequal budgets called direction test, natural signup credited, disposable identities paid, lifetime tiers farmed, one IP or device used as proof, legitimate households blocked, private abuse recipe exposed, vested rights clawed back without policy, referred-versus-organic revenue treated as causal, or spam and channel displacement omitted | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Actor and attribution policy, valid-referral and reward lifecycle, direction and equal-budget decision, tier and anti-abuse controls, commercial-partner handoff, incrementality and quality, verification, and remaining referral-incentive risk |
+| LLM token versus task or outcome pricing, bounded document image or analysis units, standard versus precision service tiers, free versus paid queue priority, first-value queue treatment, automatic restoration or bound retry rights for failed work, user-facing model disclosure or model pinning, cheap-default and premium-model packaging, automatic escalation charging, managed provider cost, BYOK, customer API keys, platform fees, included credits, margin, conversion, or retained contribution are created, changed, reviewed, or reported | `.mustflow/skills/llm-product-monetization-review/SKILL.md` | Buyer and task-unit ledger, accepted result, input and output bounds, complete internal provider cost, queue capacity fairness and abandonment, objective failure and subjective recovery classes, quote reservation and restoration identity, quality tiers and route versions, model transparency and reproducibility needs, managed and customer-funded credential contract, security provider and current authority evidence, eligible-user economics, experiments, and configured command intents | Task and outcome units, scope and overflow, quote and settlement policy, real-capacity queue classes and first-value treatment, objective-failure release or reversal, bound retry rights, model-independent tiers, execution receipts and pinning, escalation price authorization, managed and BYOK packaging, platform fee, contribution metrics, events, fixtures, tests, docs, route metadata, and synchronized templates | token cost exposed as consumer value, document or analysis unit unbounded, average cost hiding tail loss, free users intentionally slept while capacity is idle, first value queued into abandonment, strict paid-first starvation, fake countdown, copied wait target, objective failure charged, subjective dislike converted to reusable credit, retry and restoration duplicated, model name becoming unstable product contract, exact pin silently substituted, watermark treated as complete transparency, automatic escalation surprise-charged, BYOK stored client-side, customer key treated as preference, managed use secretly throttled as unlimited, free BYOK draining platform value, or per-call model price treated as accepted-result margin | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Audience and accepted result, external unit and internal cost, task bounds and quote, queue capacity fairness and wait evidence, failure classification and recovery asset, quality tiers escalation and execution disclosure, managed and BYOK economics and security, eligible-user contribution, specialist handoffs, verification, and remaining LLM monetization risk |
+| Seasonal battle passes, monthly memberships, liveops cadence, missions and reward tracks, cosmetic or power goods, convenience or sidegrades, deterministic bundles, paid or free randomized rewards, loot boxes, odds, pity, duplicates, spend concentration, production cost, fairness, retention, platform review, or regulatory risk are created, changed, reviewed, or reported | `.mustflow/skills/game-liveops-commerce-integrity-review/SKILL.md` | Team capacity, season and membership obligations, catalog classification and asset economics, fair ceiling, deterministic contents, random-reward access and odds, pity and duplicate states, age and geography authority, payer distribution, experiments, and configured command intents | Season and membership policy, missions and catch-up, catalog and power classification, cosmetic visibility, sidegrade limits, deterministic bundles, random-reward disclosure and audit, caps, metrics, fixtures, tests, docs, route metadata, and synchronized templates | impossible liveops cadence, battle pass copied from large game, daily attendance punishment, pass currency funding indefinite free seasons, membership with no ongoing value, cosmetic with competitive visibility harm, paid exclusive power, PvE content burn omitted, loot-box disclosure treated as complete safety, indirect paid randomness mislabeled free, odds or pity version missing, whale concentration hidden by ARPPU, or regulation embedded from a stale snapshot | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Team capacity and cadence, season and membership economics, catalog and fair ceiling, deterministic or random contents, odds pity duplicates age and geography, payer distribution and authority, verification, and remaining liveops-commerce risk |
+| Free-result watermarking, embedded service attribution, public-share branding, affiliate or influencer compensation, fixed CPA, first-payment or recurring commission, partner attribution and clawback, lifetime commission, owned-product cross-promotion, portfolio promotion frequency, brand relationship disclosure, channel cannibalization, incremental acquisition, or retained portfolio contribution are created, changed, reviewed, or reported | `.mustflow/skills/growth-distribution-integrity-review/SKILL.md` | Artifact ownership distribution and recipient ledger, partner audience channel qualification compensation attribution disclosure and liability ledger, source-target product fit relationship exposure suppression and data-boundary ledger, pre-exposure holdout and natural-demand evidence, full portfolio economics, current authority versions, and configured command intents | Artifact branding eligibility placement removal and recipient events, partner qualification compensation attribution disclosure fraud clawback and termination, cross-promotion eligibility relationship placement frequency suppression and data consent, causal contribution metrics, fixtures, tests, docs, route metadata, and synchronized templates | private artifact watermarked as growth engine, completed result degraded to sell removal, export count treated as reach, visible mark treated as complete AI-origin compliance, user data embedded in export, attributed gross revenue treated as incremental, branded search or coupon poaching paid, lifetime commission granted without continuing value or termination, affiliate confused with customer referral, internal inventory treated as free, source product interrupted before value, unrelated products promoted because ownership matches, shared account or data implied falsely, copied frequency cap, target clicks promoted while source or portfolio contribution falls, or current disclosure authority omitted | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Artifact distribution and branding decision, partner compensation attribution disclosure and bounded-liability decision, source-target fit relationship data frequency and suppression decision, incremental portfolio contribution and guardrails, specialist handoffs, verification, and remaining growth-distribution risk |
+| Unified versus separate product accounts, shared identity or account portals, common versus service-specific credit wallets, cross-service credit value, individual subscriptions, pure or mixed bundles, portfolio entitlements, umbrella or endorsed product brands, failure spillover, cross-product reuse, bundle cannibalization, or retained portfolio customer value are created, changed, reviewed, or reported | `.mustflow/skills/product-portfolio-integrity-review/SKILL.md` | Product customer job complement and substitute ledger, global identity and service membership ledger, wallet classes exchange cost and rights ledger, individual and bundle package migration ledger, parent and product brand fit and failure ledger, causal experiment sequence, current authority, and configured command intents | Portfolio identity and service-right policy, shared wallet classes and exchange, mixed package and entitlement policy, brand architecture and operational isolation, contribution metrics, events, fixtures, tests, docs, route metadata, and synchronized templates | unified login granting authorization, one product suspension banning the portfolio, product data merged without consent, paid and promotional lots collapsed, one nominal credit hiding cost differences, free spend called paid reuse, two purchases cannibalized by one wallet, bundle-only pricing rejecting single-product demand, multi-subscriber downgrade omitted, bundle churn hiding dormant products, common ownership treated as brand fit, logo separation called failure isolation, or all four axes changed at once | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Portfolio relationship, account and service rights, wallet and ledger handoff, package migration and economics, brand fit and isolation, experiment sequencing, retained contribution, verification, and remaining portfolio risk |
+| Single-language versus simultaneous multilingual launch, first-market language, staged locale rollout, AI or machine translation with selective human review, exploratory versus revenue-ready or fully supported locales, localized SEO demand, high-trust translation review, translation maintenance cost, per-locale conversion refunds support retention or contribution, or multilingual-growth claims are created, changed, reviewed, or reported | `.mustflow/skills/localization-market-expansion-review/SKILL.md` | Buyer and market demand ledger, locale stages and supported surfaces, translation source model glossary risk and review ledger, localized SEO query URL and alternate mapping ledger, language-funnel economics, experiment or rollout evidence, current search store payment and jurisdiction authority, and configured command intents | Locale qualification and promotion policy, support-depth matrix, risk-based human review, AI translation versioning, honest product-language disclosure, international SEO demand test, contribution metrics, events, fixtures, tests, docs, route metadata, and synchronized templates | English assumed universal first language, code readiness treated as language launch, language count as success, many locales opened before paid learning, machine translation accepted for rights-bearing copy, review assigned only by language rather than consequence, literal keyword translation called SEO, translated page hiding an untranslated product, reciprocal locale mapping absent, scaled low-value pages, traffic called revenue, recurring maintenance omitted, or silent fallback after selling support | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | First language and market, locale stage and surface ownership, translation and review evidence, SEO usefulness and mapping, product-language disclosure, full-funnel contribution, promotion or retirement decision, verification, and remaining locale-expansion risk |
+| Direct website versus Microsoft Store, Mac App Store, another approved desktop storefront, or hybrid desktop distribution; store versus external commerce; channel fees, install trust, signing, notarization, sandbox and capability fit, discovery, checkout, trials, licenses, restoration, entitlement portability, update ownership, review delay, enterprise or offline distribution, support, or retained channel contribution are created, changed, reviewed, or reported | `.mustflow/skills/desktop-commercial-distribution-review/SKILL.md` | Product buyer platform price capability privilege enterprise and offline ledger, current channel policy and region ledger, complete channel economics, transaction receipt license restore and migration ledger, common core and channel adapter ledger, update and support ownership, experiment evidence, and configured command intents | Direct store and hybrid channel matrix, capability and sandbox decision, commerce and license adapters, entitlement provenance restoration and portability, update ownership, economics and break-even lift, causal channel metrics, fixtures, tests, docs, route metadata, and synchronized templates | storefront presence called incremental discovery, nominal fee treated as total cost, stale game or regional policy applied to another app, trusted store build silently crippled, direct binary trust cost omitted, product core forked by channel, store login used as only customer identity, duplicate purchase or restore missing, portability promised without contract, review delay and urgent fix path omitted, downloads called conversion, or self-selected channel cohorts treated as causal | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Product and channel fit, current policy and capabilities, discovery delivery commerce entitlement update and support ownership, complete economics, core and adapters, restore and migration, contribution evidence, verification, and remaining desktop-distribution risk |
+| Responsive web-first versus mobile app-first, mobile web versus PWA or native clients, operating-system capability dependence, native break-even users, app acquisition versus retention, or local-first versus cloud-first data ownership, offline work, synchronization, privacy, encryption, backup, and recovery are created, changed, reviewed, or reported | `.mustflow/skills/client-platform-strategy-review/SKILL.md` | Product job and first-value ledger, OS capability matrix, comparable full funnel, client lifecycle cost, data authority and size ledger, sync conflict and recovery ledger, current platform policy, experiment evidence, and configured command intents | Web PWA cross-platform and native sequencing, capability matrix, qualified CAC and retained contribution, break-even sensitivity, self-selection controls, data authority, offline sync conflict privacy encryption backup and recovery policy, fixtures, tests, docs, route metadata, and synchronized templates | platform prestige, app install called acquisition success, raw app retention compared with web visitors, PWA label driving premature offline complexity, copied MAU break-even, nonpositive per-user denominator hidden, OS capability assumed from another platform, local-first chosen to save raw storage cost, cloud-first treated as always online, last-write-wins data loss, CRDT treated as product policy, local storage called private without encryption or recovery, or server authority collapsed into user-content authority | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Product and platform job, capability and launch sequence, comparable funnel economics and break-even range, data authority and sync semantics, privacy and recovery, current source boundary, verification, and remaining client-platform risk |
+| Shared versus independent backend or platform investment, common-core break-even project count, correlated outage and recovery congestion, service cells, effective service count, low-usage service continuation, rescue, archive or shutdown, search and cross-product preservation, closure NPV, required growth, founder opportunity cost, new-service launch versus existing-product conversion or retention improvement, marginal contribution, diminishing returns, context switching, or portfolio capacity are created, changed, reviewed, or reported | `.mustflow/skills/service-portfolio-capital-allocation-review/SKILL.md` | Portfolio and lifecycle ledger, shared-platform fixed residual and de-sharing costs, independent comparator, correlated failure and recovery ledger, per-service contribution preservation closure and growth ledger, work-candidate expected value, founder capacity, current obligations, and configured command intents | Failure-adjusted shared-platform break-even interval, cell topology, per-service continue rescue archive sell or shutdown decision, preservation and closure plan, required-growth and experiment gates, marginal time allocation, active work set, fixtures, tests, docs, route metadata, and synchronized templates | raw project count as platform law, reusable code equated with shared database, correlated failure omitted, more services assumed always better, MAU used as shutdown threshold, gross revenue replacing contribution, weighted scorecard hiding units, search and cross-product value double-counted, founder time priced at nominal wage, sunk cost earning another rescue, one noisy month triggering closure, fixed launch-improve percentage, PMF-free conversion polishing, context switching omitted, or customer export refund notice and deletion obligations ignored | `changes_status`, `changes_diff_summary`, `lint`, `build`, `test_related`, `test`, `docs_validate_fast`, `test_release`, `mustflow_check` | Portfolio objective and capacity, shared-independent economics and failure interval, isolation cells, per-service continuation and closure value, rescue and obligation gates, launch-improvement marginal allocation, verification, and remaining portfolio-allocation risk |
 | Repository improvement, audit, prioritization, stabilization, polish, onboarding, contributor-readiness, production-readiness, or iterative improvement is requested without a single predetermined edit | `.mustflow/skills/repo-improvement-loop/SKILL.md` | User goal, improvement mode, repository evidence, candidate risks, current changed files, and command contract entries | Repository diagnosis, ranked candidates, and at most one scoped improvement cycle unless the user explicitly requests analysis-only | idea spam, ungrounded prioritization, autonomous loop drift, broad rewrite, or unverified improvement claim | `changes_status`, `changes_diff_summary`, `docs_validate_fast`, `test_release`, `mustflow_check` | Mode, evidence inspected, scored candidates, selected improvement, files changed or analysis-only note, verification, next improvement question, and stop reason |
 | Current repository evidence reveals a scope-adjacent bug, missing test, stale synchronized surface, public-contract drift, security or privacy exposure, data-loss risk, brittle error handling, concurrency risk, operational risk, or UX inconsistency outside the literal request | `.mustflow/skills/proactive-risk-surfacing/SKILL.md` | Literal user request, current evidence, risk relationship, severity, expected edit size, authority boundary, and verification options | Fix-or-report decision, small related fixes, focused tests or synchronized surfaces, and final proactive risk notes | scope creep, speculative cleanup, hidden broad refactor, ignored high-severity risk, or false completion claim | `changes_status`, `changes_diff_summary`, `test_related`, `docs_validate_fast`, `test_release`, `mustflow_check` | Candidate decisions: fix now, report only, ask first, or ignore; files changed, verification, and remaining proactive risks |
 | A final report or completion claim needs current evidence for changed files, requirements, command receipts, skipped checks, synchronized surfaces, repository-local verification boundaries, or remaining risks | `.mustflow/skills/completion-evidence-gate/SKILL.md` | User goal, selected repository and command contract, changed-file evidence, explicit parent dependencies, skills used, verification results, skipped checks, synchronized surfaces, and remaining risks | Final report evidence and the smallest missing in-scope evidence surface only | false completion, stale receipts, hidden skipped checks, cross-root verification bleed, unrelated parent blocker, unsupported readiness claim, or contract drift | `changes_status`, `changes_diff_summary`, `test_related`, `test`, `test_audit`, `lint`, `build`, `docs_validate_fast`, `docs_validate`, `test_release`, `mustflow_check` | Completion status, repository verification boundary, requirement evidence map, changed and synchronized surfaces, commands run, skipped checks, and final wording boundary |