Architecture Review

    2

    Runs an evidence-based architecture review: criteria, coverage, findings, owner responses and a recommended disposition.

    Free

    4 usesSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Architecture Review

    Example session with this skill installed

    Run a formal architecture review of our payment gateway redesign before it goes to production.

    • Subject: payment-gateway v2.4.0-rc1, specification PAY-ARCH-2026-v2 (excerpt below).
    • Decider: Enterprise Architecture Board (EAB). Reviewer: me, platform architect, no conflicts.
    • Criteria: POL-SEC-04 §3.2 "OAuth access and refresh tokens must be encrypted at rest in every cache, queue or datastore that holds them", and NFR-PAY-07 "p99 authorization latency ≤ 250 ms at 400 TPS".
    • Spec §5.3: tokens are cached in the Redis 7 cluster redis-pay (3 primaries, 3 replicas), TTL 900 s. TLS is disabled inside the VPC. RDB snapshots go to s3://pay-backups daily.
    • Load test LT-118 on rc1, staging, 30 minutes: p99 212 ms at 400 TPS.
    • Payments team lead: "Redis is network-isolated, so encryption is not needed."

    Write the report under docs/.

    • Read your context and instructions
    • Compiled the architecture review

    Wrote docs/architecture/tasks/pay-gw-review-001/arch-review-task/architecture-review-report.md. One confirmed nonconformity, one criterion met within tested scope, three evidence gaps. The disposition below is a recommendation for the EAB, not an approval.

    Architecture Review Report: payment-gateway v2.4.0-rc1 [ARR-PAY-001]

    Summary

    Review of payment-gateway v2.4.0-rc1 against PAY-ARCH-2026-v2 for the EAB production-release decision. Two criteria were examined.

    FD-1 is a confirmed nonconformity: OAuth tokens are held unencrypted at rest in redis-pay and copied into daily RDB snapshots. The latency criterion is met in staging (FD-2). Three gaps (G-1 to G-3) limit what this review can say. Recommended disposition:

    conditional acceptance, conditions C-1 to C-3. The EAB decides.

    Review Subject and Mandate [RS-1]

    FieldValueClassificationSource
    Subjectpayment-gateway v2.4.0-rc1providedrequest
    BaselinePAY-ARCH-2026-v2, §5.3 excerptprovidedrequest
    Decision useproduction releaseprovidedrequest
    DeciderEnterprise Architecture Boardprovidedrequest
    Reviewerplatform architect, declared no conflictsprovidedrequest
    Excludedcode review, penetration test, operational readinessderivedreview boundary

    Criteria

    CR-1, POL-SEC-04 §3.2. Question: are access and refresh tokens encrypted at rest in every store that holds them? Oracle: each such store (primary memory, replicas, snapshots, backups) shows an at-rest encryption control.

    CR-2, NFR-PAY-07. Question: is p99 authorization latency ≤ 250 ms at 400 TPS? Oracle: a load test on the exact subject revision.

    No other criterion was applied. Severity is not assigned: no severity scale was supplied.

    Evidence and Coverage

    IDEvidenceRevisionMethodCoversLimits
    EV-1Spec §5.3 excerptPAY-ARCH-2026-v2document inspectionCR-1excerpt only, full spec not supplied
    EV-2Load test LT-118rc1staging load test, 30 minCR-2staging only, single run, payload mix unknown
    EV-3Team lead statement—owner responseCR-1assertion, no configuration evidence

    Covered: CR-1 for Redis primaries, replicas and RDB snapshots; CR-2 in staging. Not covered: other stores that may hold tokens, key management, production latency.

    Findings

    FD-1: confirmed nonconformity (CR-1)

    • Location: spec §5.3, token cache in redis-pay; RDB snapshots in s3://pay-backups.
    • Condition: tokens are cached in Redis with no at-rest encryption stated, TLS disabled, and copied to S3 by daily snapshots.
    • Impact: anyone who can read replica memory, snapshot files or the backup bucket can read live tokens (TTL 900 s in the cache; retention of snapshots not supplied).
    • Owner response: disputed. "Redis is network-isolated."
    • Status: confirmed. Network isolation limits who can reach the cluster; POL-SEC-04 §3.2 requires encryption at rest, a different control. Whether isolation can stand in for encryption is an exception decision for the policy owner, not a reviewer call.
    • Confidence: high for Redis (explicit in EV-1); medium for S3, because bucket encryption was not supplied (G-2).

    FD-2: observation, criterion met within tested scope (CR-2)
    LT-118 shows p99 212 ms at 400 TPS on rc1, below the 250 ms limit. This covers one 30-minute staging run; it says nothing about production traffic.

    Gaps

    • G-1: no inventory of other components that hold tokens (logs, queues, analytics exports).
    • G-2: encryption setting of s3://pay-backups not supplied.
    • G-3: owner of any key-management control not named.

    Recommended Disposition for the EAB [DP-1]

    Conditional acceptance, if the EAB agrees:

    C-1 Evidence that tokens are encrypted at rest in redis-pay, either by the store or by the application before caching, verified on rc1 or a later revision.

    • C-2 Evidence that s3://pay-backups encrypts snapshots at rest.
    • C-3 A token-store inventory signed by the subject owner, closing G-1.

    Alternative: a time-bound exception to POL-SEC-04 §3.2 granted by its policy owner, with scope, compensating controls and expiry.

    Invalidation: any change to token storage, or a new subject revision, reopens CR-1.

    Handoffs

    ToNeedsAccepted when
    Payments team lead (subject owner)G-1 inventory, G-2 bucket configurationinventory signed; bucket encryption configuration exported
    POL-SEC-04 policy ownerexception decision, if requestedsigned exception with scope, mitigation and expiry, or a refusal
    Security architecture / application engineeringhow to encrypt tokensdesign accepted by its owner (outside this review)

    Explicit Unknowns

    Other token stores (G-1). Backup bucket encryption (G-2). Key-management owner (G-3). Production latency. Snapshot retention period.

    Traceability

    ClaimClassificationSourceFreshness
    Tokens cached in Redis, TLS offprovidedEV-1as supplied 2026-09-15
    Policy requires encryption at restprovidedPOL-SEC-04 §3.2, quoted in requestas supplied
    Isolation is not encryption at restderivedCR-1 wording against EV-3—
    p99 212 ms at 400 TPSprovidedLT-118run date not supplied
    DP-1recommendationthis reviewpending EAB decision

    Verification

    No validator was supplied, so no command was run. Reviewer self-checks, which support but do not replace the EAB decision:

    • every finding traces to a criterion and evidence (FD-1: CR-1, EV-1, EV-3; FD-2: CR-2, EV-2);
    • only the two supplied criteria were applied;
    • the reviewer granted no approval.

    Next steps

    1. Send G-1 and G-2 requests to the payments team lead.
    2. Put DP-1 on the EAB agenda with this report.
    3. If the team pursues isolation instead of encryption, route the exception request to the POL-SEC-04 owner.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Architecture Review: Full Description

    What it does

    This skill performs one evidence-bounded review of an exact architecture subject against authoritative concerns and criteria. It records coverage, findings, limitations, responses and governing disposition without choosing architecture or granting itself approval authority.

    Use it when

    Use when an exact architecture proposal/specification/baseline/change needs review for an identified decision under a supplied mandate.

    For example: “We need formal sign-off on the core payment gateway redesign before going to production. Security says the auth token handling violates enterprise policy.”

    What you get

    • Architecture Review Report

    Written as Markdown to <your output folder>/architecture/tasks/<run-id>/arch-review-task/.

    What it will not do

    Do not use for architecture design/advice, portfolio-level review governance, code/PR/security/operational-readiness review, compliance audit/certification, checklist execution or meeting facilitation.

    How it works

    1. Check an architecture review is required.
    2. Bound the review subject and mandate.
    3. Freeze criteria and oracles.
    4. Evaluate evidence against criteria.
    5. Form findings and obtain responses.
    6. Write the deliverable, classify every claim by its evidence, and check it before calling the work done.

    What's in the package

    Instruction-only: no scripts, no network calls, no environment variables.

    • LICENSE.txt
    • SKILL.md
    • agents/openai.yaml
    • assets/output-template-task.md
    • references/domain-rules.md
    • references/operating-rules.md
    • references/output-contract.md

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.

    • Version: 1.0
    • License: proprietary, LICENSE.txt in the zip
    • Also sold in: Delivery, Testing and Governance Pack, Architect Complete Suite

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    4 uses

    Times used to date

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Permissions

    Write Files
    Read Files
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    No. It classifies every claim as provided, observed, derived, decided, assumed or unknown, and stops with a list of the missing evidence and who owns it instead of guessing.

    No. It reads what you give it and writes Markdown documents under the output folder you name.

    Any agent that supports SKILL.md skills, including Claude Code, Codex CLI, Cursor and Gemini CLI. It has no scripts and makes no network calls.

    New in Code Quality & Review

    Free