- Home
- Skills
- Code Quality & Review
- Change Assurance Review
Works with the AI Agents you already use
Change Assurance Review
Review a concrete change against stated intent, project invariants, contracts, and regression risk, returning severity-calibrated findings and explicit unverified surfaces.
$15
Change Assurance Review
Example session with this skill installed
Review PR #412 against main: it adds GET /api/users/:id returning the full user row. Check intent, authorization, response shape, regression evidence, and report severity-calibrated findings with verified and unverified surfaces.
- Read your context and instructions
- Compiled the change assurance review
Illustrative worked example. Synthetic data from the package; this is not an observed external execution.
TARGET / BASELINE: PR #412 vs main; files: routes/users.js, no tests added.
FINDINGS
- BLOCKER | routes/users.js:22 | no authorization check; any authenticated user can read any user's row (IDOR) | cross-user data exposure | scope query to session user or enforce role
- MAJOR | routes/users.js:24 | returns password_hash and reset_token fields | sensitive data exposure | select explicit safe columns
- NOTE | no test for the new endpoint | regression risk | add authz + shape test
VERIFIED: route registered correctly; input id validated as integer.
UNVERIFIED: production auth middleware order (not in diff).
RESIDUAL RISK: medium until authz + field filtering added.
VERDICT: BLOCKED
Connects securely to your tools. The creator never sees your data.
About this skill
When a concrete diff, pull request, configuration change, or document must be checked before merge, Change Assurance Review pins the exact target and baseline, reads the smallest sufficient evidence, and checks the change against stated intent, project invariants, contracts, and regression risk. It returns severity-calibrated findings while separating verified evidence from unverified surfaces.
What it does
It reviews the change that exists, not an imagined future system. Project governance and accepted specifications outrank committed contracts, current repository evidence, this workflow, and generic preferences. The review stays bounded to the named target, baseline, requested intent, relevant architecture, and reachable tests or evidence.
How it works
- Pin the exact diff, files, artifact, comparison baseline, requested intent or specification, and relevant evidence.
- Apply the intent lens to acceptance criteria and wrong or missing edge behavior.
- Apply the integrity lens to invariants, public contracts, domain semantics, error paths, data handling, dependency direction, and compatibility.
- Apply the risk lens to regression surface, security-sensitive changes, migrations or configuration, concurrency or state transitions, observability, and rollback implications when applicable.
- Report findings with severity, target evidence, violated intent or invariant, consequence, and correction direction.
Output
The review returns:
- TARGET / BASELINE
- FINDINGS
- VERIFIED
- UNVERIFIED
- RESIDUAL RISK
- VERDICT
The verdict uses severity-calibrated semantics. BLOCKED is reserved for a blocker-level correctness, security, data-loss, or contract issue, or missing evidence for a critical invariant. UNVERIFIED is reported separately and is not treated as a pass.
Boundaries
This is a review workflow for an existing concrete change. It does not choose architecture, resolve an unfinished design, provide a final release-readiness verdict, modify the implementation, install dependencies, commit, push, merge, deploy, or run destructive migrations. If required evidence or capability is unavailable, it records UNVERIFIED or BLOCKED instead of fabricating a result.
Why it is different
An architecture or decision workflow resolves what should be built, while a final readiness workflow certifies a downstream release state. Change Assurance Review starts with the change that exists and gives a bounded evidence-backed review handoff against intent, invariants, contracts, and regression risk.
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.
This is a review workflow, not architecture selection, unresolved-design resolution, or final release readiness. It requires a concrete diff, pull request, configuration change, or document, a comparison baseline, stated intent or specification, and relevant evidence. It is read-only and does not modify code, install dependencies, commit, push, merge, deploy, or run destructive migrations. Shell, Git, browser, network, MCP, and package-manager capabilities are not assumed; unavailable checks are marked UNVERIFIED or BLOCKED.
How to use
- 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
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 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
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
Permissions
Requires read access to the supplied diff, target files, requested intent or specification, and relevant project evidence. Terminal/Shell is optional for focused read-only checks; if unavailable, use repository-native or manual checks and mark affected items UNVERIFIED or BLOCKED. Browser, network access, MCP services, external accounts, write access and environment variables are not required by the package.