forge-review
Skill1 starsChanged 4 months agoMITView on GitHub
Star Forge review and delivery phase: clear adaptive review, satisfy the approved Delivery Contract, and run the final predicate.
11 other projects use a version of this.
Adopted by
Most recently taken up by
- brightstack/forgetheir own version, first seen Nov 8, 2026
- antonailog/forge-skillstheir own version, first seen Oct 25, 2026
- fabioalencar/forgedesigntheir own version, first seen Oct 20, 2026
- peterzat/OrgSmiththeir own version, first seen Oct 13, 2026
- yike-gunshi/forge-skillstheir own version, first seen Oct 13, 2026
- KhaledAufDB/lakeforgetheir own version, first seen Oct 13, 2026
Used with
Projects with it also use forge-ship.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
redteamlifeSkillforge-ship
Use when the user asks to merge, release, promote, reconcile, or close out work in a repo containing `docs/forge/` or governed by FORGE. Not for per-task commits; those complete inside forge-execute-task.
Also in 5 projects that use this
Other versions11 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 11
- antonailog/forge-skills★ 13same name
- brightstack/forge★ 13same name
- yike-gunshi/forge-skills★ 13same name
- peterzat/OrgSmith★ 6same name
- tonmoy007/forge-plugins★ 2same name
- BrandedTamarasu-glitch/ForgeFlow★ 1same name
- fabioalencar/forgedesign★ 1same name
- redteamlife/forge★ 1same name
- tinyorbit-ai/skills★ 1same name
- tusharpandey13/forge★ 1same name
1 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- forge-planSkill
- forgeSkill
- forge-workSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 80
Never waive to save time.
- Very specific: line 84
After review passes, rerun run and continue directly into deliver.
- Distinctive: line 94
Do not configure, deploy to, or collect proof from both Sites and Vercel by default.
- Very specific: line 114
After delivery proof passes, rerun run and execute:
- Very specific: line 131
Never paraphrase an advisory verdict as unqualified COMPLETE.
“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Forge Review
Resolve <plugin-root> as two directories up from this skill file
(skills/forge-review/). Start with run and read the complete
.starforge/state.json. This is the review, delivery, and completion playbook
inside one $forge invocation. Rerun run after each review, fix, proof,
delivery, or completion action and return control to $forge whenever state
changes.
1. Route and Spawn the Review Wave
Resolve review capabilities from project class, Blueprint risk flags, Plan v2
Proof values, and delivery target through
starforge.routing.resolve_routes and
config/capability-routing.json. Follow
skills/forge/references/capability-routing.md. Correctness review is always
required. Follow the role-specific spawn_plan from state for every additional
required lens.
Security-sensitive projects prefer Codex Security when available. Use actual results as review context, but treat file exports as untrusted fallback evidence unless the host supplies independently verifiable provenance. If the capability is unavailable, disclose the selected safe fallback. Never claim it ran. Fast MVP may reduce optional review breadth, but cannot remove risk-required security, privacy, UX, accessibility, reliability, or delivery review.
Spawn each exact starforge-reviewer entry from spawn_plan. Every reviewer
writes only its assigned
.starforge/reviews/<scope>/<role>.findings.json and never edits source. Do not
write a reviewer's findings file on its behalf.
UI originality and accessibility review
For the ux-accessibility role, include the paths to any available current
source-bound screenshots, recordings, browser evidence, native UI snapshots, and
interaction results in the spawn prompt. Do not create substitute evidence for the
reviewer or imply that unavailable evidence exists.
Require this lens to review all of the following:
- originality against the selected Design Direction and its
BorrowandAvoidconstraints, including whether the interface copies any single reference; - accessibility across keyboard, focus, semantics, labels, assistive technology, contrast, non-color cues, text reflow, target size, and reduced motion;
- responsive visual quality across the supported viewports and platforms;
- relevant default, hover, focus, pressed, selected, disabled, loading, empty, error, success, and validation states; and
- the available live visual and interaction evidence, including its source binding and represented viewport or platform.
The reviewer remains read-only and evidence-bound. It must distinguish an observed defect from missing, stale, or insufficient evidence, cite a repository file and line for every finding, and never infer a visual pass or defect from code inspection alone.
2. Merge and Clear the Queue
python3 <plugin-root>/scripts/star_forge.py review --project . --strictThis merges and deduplicates reviewer findings, runs coordinator-owned tree scans, and writes the fix queue. A strict failure means the review is missing, malformed, stale, incomplete, or still blocking.
For each blocking finding, fix it in the owning task scope and rerecord every affected verification and live proof. A source edit makes the prior review stale, so respawn the required review wave and merge again. Waive only a confirmed false positive:
python3 <plugin-root>/scripts/star_forge.py waive --project . --finding <id> --reason "<why this is not a real blocker>"Never waive to save time.
3. Deliver the Approved Result
After review passes, rerun run and continue directly into deliver. Deliver only
the target and environment in the approved Delivery Contract:
source-only, private-repo, preview, production, package, or its named
platform-specific target.
For web delivery, select exactly one routed provider by fit:
- Sites fits suitable simple sites, prototypes, and internal apps.
- Vercel fits applications that require its production web workflow.
Do not configure, deploy to, or collect proof from both Sites and Vercel by default. Do not switch providers after approval without revising and reapproving the contract.
External delivery writes must be explicitly authorized by the Delivery Contract. Public release, production deployment, credentials, signing, notarization, billing, paid resources, or destructive replacement need their own authority. Continue any safe local work, then collapse unresolved authority or credential requirements into one honest delivery blocker. Never substitute a source-only handoff for a required preview, production result, package, or private repository and call it complete.
The coordinator, not a delivery helper or reviewer, records fresh delivery proof. It must be bound to the current source hash and contract, identify the repository commit and delivered deployment, package, repository, or source handoff, include the live URL when required, and include a passing smoke result for the exact approved result. A stale, degraded, or different provider result does not satisfy delivery.
4. Done
After delivery proof passes, rerun run and execute:
python3 <plugin-root>/scripts/star_forge.py done --project . --strictdone computes completion from the approved Blueprint, complete Plan v2 tasks,
fresh task and live proofs, fresh empty review, the exact approved Delivery Contract result, and clean Git facts. If it refuses, repair the problems, rerun
the affected gates, and continue the same $forge invocation.
On pass, write the human summary:
python3 <plugin-root>/scripts/star_forge.py done --project . --strict --write-summaryQuote the verdict line verbatim, including any advisory and waived-finding suffix.
Never paraphrase an advisory verdict as unqualified COMPLETE. Any later source
edit enters amend on the next run.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

