Kiro Review logo

Kiro Review

CommunityPopular
gotalab
kiro-review

Review a task implementation against approved specs, task boundaries, and verification evidence. Use after an implementer finishes a task, after remediation, or before accepting a task as complete.

Overview

Publishergotalab
Repositorycc-sdd
Skill namekiro-review
Stars
3.7K
Forks
283
Bundled files
1
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 1 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by gotalab on GitHub. Read the source before you install it.

Installation

Install the Kiro Review AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/gotalab/cc-sdd.git /tmp/cc-sdd
mkdir -p .claude/skills
cp -r /tmp/cc-sdd/tools/cc-sdd/templates/agents/antigravity-skills/skills/kiro-review .claude/skills/kiro-review
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Kiro Review in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Kiro Review on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Kiro Review is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

kiro-review

<background_information> This skill performs task-local adversarial review. It verifies that the implementation is real, complete, bounded, aligned with approved requirements and design, and supported by mechanical verification evidence.

Boundary terminology continuity:

  • discovery identifies Boundary Candidates
  • design fixes Boundary Commitments
  • tasks constrain execution with _Boundary:_
  • review rejects concrete Boundary Violations </background_information>
  • After an implementer reports READY_FOR_REVIEW
  • After remediation for a rejected review
  • Before marking a task [x]
  • Before accepting a task into feature-level validation

Do not use this skill to invent missing requirements or silently reinterpret the spec.

Inputs

Provide:

  • Task ID and exact task text from tasks.md
  • Relevant requirement section numbers
  • Relevant design section numbers
  • Spec file paths (requirements.md, design.md, optionally tasks.md)
  • The implementer's status report
  • The task _Boundary:_ scope constraints
  • Validation commands discovered by the controller
  • Relevant steering excerpts when applicable
  • Relevant ## Implementation Notes entries when applicable

Outputs

Return one of:

  • APPROVED
  • REJECTED

Also return:

  • Mechanical results
  • Findings with severity
  • Required remediation
  • One-sentence summary

Use the language specified in spec.json.

First Action

Run git diff to inspect the actual code changes. If the diff is large or ambiguous, read the changed files directly. Do not trust the implementer report as source of truth.

Core Principle

Read the spec yourself. Read the diff yourself. Verify mechanically where possible. Reject on concrete failures rather than interpretive optimism. The main review question is not just "does it work?" but "does it stay inside the approved responsibility boundary without hiding new coupling?"

Mechanical Checks

Run these checks and use the result as primary signal.

1. Regression Safety

  • Run the project's canonical test suite using the validation commands discovered by the controller.
  • If tests fail, reject.

2. No Residual Placeholder Markers

  • Check changed files for TBD, TODO, FIXME, HACK, XXX.
  • Reject if new placeholder markers were introduced without explicit task justification.

3. No Hardcoded Secrets

  • Check changed files for hardcoded secrets or credentials.
  • Reject if concrete secret patterns are introduced.

4. Boundary Respect

  • Compare changed files against the task _Boundary:_ scope.
  • Reject if the change spills outside the approved boundary without explicit justification.
  • Reject if the implementation introduces hidden cross-boundary coordination inside what should be a local task.

5. RED Phase Evidence

  • For behavioral tasks, verify that the implementer status report includes RED_PHASE_OUTPUT.
  • Reject if RED evidence is missing, empty, or unrelated to the task's acceptance criteria.

6. Runtime-Sensitive Static Checks

  • If the project already has lint or equivalent static analysis for the touched stack, run the relevant command for the task boundary.
  • Pay attention to patterns that can survive typecheck/build yet fail at runtime: type-only imports used as values, missing namespace value imports for qualified-name access, unresolved globals, and newly introduced runtime-sensitive dependencies without matching boot/runtime handling.
  • If no project lint command exists, perform a targeted diff-based spot check in the changed files for those patterns.
  • Reject on concrete findings that create a realistic boot-time or module-load failure.

Judgment Checks

7. Reality Check

  • Confirm the implementation is real production code, not a placeholder, stub, fake path, or deferred-work shell.

8. Acceptance Criteria Coverage

  • Read the task description and confirm all aspects are implemented, not only the primary happy path.

9. Requirements Alignment

  • Read the referenced sections in requirements.md.
  • Confirm each requirement is satisfied by concrete observable behavior.
  • Use original section numbers only.

10. Design Alignment

  • Read the referenced sections in design.md.
  • Confirm the implementation uses the prescribed structures, interfaces, and dependency direction.
  • Reject silent substitutions for design-mandated choices.

10.5 Boundary Audit

  • Compare the implementation against the design's boundary commitments and out-of-boundary statements.
  • Reject if downstream-specific behavior is pushed into an upstream boundary for convenience.
  • Reject if the implementation creates new hidden dependencies, shared ownership, or undeclared coupling across adjacent boundaries.
  • Reject if a task that is not an explicit integration task now behaves like one.

11. Test Quality

  • Confirm tests prove the required behavior rather than only scaffolding.
  • Confirm tests would fail if the implementation were removed or broken.

12. Error Handling

  • Confirm relevant failure paths are handled and not silently swallowed.

Severity Model

Use:

  • Critical for broken functionality, invalid verification, data loss, security risk, or major scope violation
  • Important for required fixes before acceptance
  • Suggestion for non-blocking improvements
  • FYI for informational notes

Stop / Escalate

Escalate instead of papering over the issue when:

  • The approved spec is ambiguous in a correctness-critical way
  • The design conflicts with what is technically possible
  • Required evidence cannot be gathered
  • The implementation only works by silently deviating from approved scope
  • Boundary ownership cannot be determined cleanly from requirements, design, and task scope

Common Rationalizations

RationalizationReality
“Tests pass, so approve”Passing tests do not prove spec compliance or boundary respect.
“The extra behavior is useful”Extra behavior outside approved scope is still drift.
“The implementer said RED was done”RED must be evidenced, not asserted.
“This gap is small enough to let through”Real gaps must be rejected or escalated.

Output Format

md
## Review Verdict
- VERDICT: APPROVED | REJECTED
- TASK: <task-id>
- MECHANICAL_RESULTS:
  - Tests: PASS | FAIL (command and exit code)
  - TBD/TODO grep: CLEAN | <count> matches
  - Secrets grep: CLEAN | <count> matches
  - Static checks: PASS | FAIL | SPOT_CHECKED
  - Boundary: WITHIN | <files outside boundary>
  - Boundary audit: CLEAN | <spillover / hidden dependency findings>
  - RED phase: VERIFIED | MISSING | N/A
- FINDINGS:
  1. <specific finding with exact files/spec refs>
- REMEDIATION: <mandatory if REJECTED>
- SUMMARY: <one sentence>

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Kiro Review AI skill do?

Review a task implementation against approved specs, task boundaries, and verification evidence. Use after an implementer finishes a task, after remediation, or before accepting a task as complete.

Why use Kiro Review on TypingMind?

Because you install it once and use it with any model. Kiro Review is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Kiro Review in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/gotalab/cc-sdd/tree/main/tools/cc-sdd/templates/agents/antigravity-skills/skills/kiro-review. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Kiro Review?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Kiro Review?

As many as you like. As long as a model supports skills, you can use Kiro Review with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Kiro Review AI skill free?

Yes. It is published on GitHub by gotalab under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇