From ironforge
Reviews implemented code via agency-agents before merge. Runs @engineering-code-reviewer on the diff, then @testing-reality-checker for QA sign-off. Blocks on critical findings and loops back to /ironforge:implement for fixes.
npx claudepluginhub romaindecoster/ironforge --plugin ironforgeThis skill is limited to using the following tools:
You are executing the Ironforge review phase. Run two passes — code review then QA sign-off.
Generates design tokens/docs from CSS/Tailwind/styled-components codebases, audits visual consistency across 10 dimensions, detects AI slop in UI.
Records polished WebM UI demo videos of web apps using Playwright with cursor overlay, natural pacing, and three-phase scripting. Activates for demo, walkthrough, screen recording, or tutorial requests.
Delivers idiomatic Kotlin patterns for null safety, immutability, sealed classes, coroutines, Flows, extensions, DSL builders, and Gradle DSL. Use when writing, reviewing, refactoring, or designing Kotlin code.
You are executing the Ironforge review phase. Run two passes — code review then QA sign-off. Block on critical findings. Do not approve a merge with unresolved blocking issues.
Get the changes to review:
# All commits since branching from main
git diff main...HEAD 2>&1
# Or if on main, commits since last tag
git diff $(git describe --tags --abbrev=0)...HEAD 2>&1
Also get the list of files changed:
git diff --name-only main...HEAD 2>&1
If the diff is empty, stop and tell the user there is nothing to review.
Present a brief summary of what changed (files touched, rough scope) before activating agents.
If graphify-out/ exists in the project root, invoke the ironforge:graph-explore skill before activating the reviewer:
Skill({ skill: "ironforge:graph-explore", args: "Qui appelle les modules modifiés ? Quel est le blast radius du diff ?" })
Use the returned context to enrich the review in Step 3: flag any caller or dependent module that may be impacted but is not covered by the diff or existing tests.
If graphify-out/ is absent or ironforge:graph-explore returns empty, skip this step and proceed normally.
Activate @engineering-code-reviewer.
Provide the agent with:
.sudocode/) for acceptance criteria contextdocs/arch/architecture.md) if it existsThe agent should evaluate:
Correctness
Code quality
Security
For each finding, the agent must classify:
Present the full review output.
If there are BLOCKING findings:
List them clearly:
## Blocking findings
1. [file:line] Description of the issue
Fix: suggested correction
2. ...
Then stop and tell the user:
These findings must be fixed before proceeding. Run
/ironforge:implementto address them, then re-run/ironforge:review.
Do not continue to Step 5 until all BLOCKING findings are resolved.
If there are only SUGGESTION and NITPICK findings:
Present them and ask the user which (if any) they want to address. Do not block on them. Continue to Step 5.
If there are no findings:
Note it and continue to Step 5.
Activate @testing-reality-checker.
Provide the agent with:
The agent should verify:
Test coverage
Test quality
Reality check
For each gap, classify as BLOCKING or SUGGESTION using the same criteria as Step 3.
Present the full QA output.
Same logic as Step 4: block on BLOCKING, present suggestions without blocking.
If all findings are resolved or non-blocking:
## Review — Approved
- Code review: ✓
- QA sign-off: ✓
**Suggestions noted:** X (non-blocking)
Ready to merge.
If blocking findings remain unresolved:
## Review — Blocked
Fix the blocking findings above, then re-run /ironforge:review.