review-work
Review Workflow implementation or corrections against its plan, read-only.
Skill by geldmacher
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/geldmacher-review-work
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/review-work
What's inside
- Skill name
review-work- Description
- Review Workflow implementation or corrections against its plan, read-only.
- License
- MIT
- Instructions
- 19 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# review-work Read the [working agreement](../../references/workflow.md) unless it is already in this task's context. Use for a requested assessment of implementation or corrections, not merely an explanation of existing findings or a progress summary. Identify the applicable plan, implementation or correction report, and current working state. When that plan records another branch or worktree, read [concurrent work](../../references/concurrent-work.md) and review that checkout. A delegated Auto-Work review follows the same rubric; return its result to the executor without requiring the human to change host modes. Remain repository-read-only, including tracked and untracked files. Select inspections and tests whose effects are permitted; put any permitted transient outputs outside the repository. If a necessary check would mutate the repository or cross another permission boundary, record the missing evidence. Do not fix code, tests, documents, or verifiers during Review. Assess every success criterion and relevant boundary against actual work; distinguish task changes from unrelated edits. Start with the smallest sufficient inspection. Before reusing a check result, inspect its origin, actual output, coverage, and applicability to current source, dependencies, configuration, and relevant environment. Explain the reuse basis briefly. A command list or success summary alone is insufficient. Reviewing evidence does not require rerunning every check. Recheck affected or uncertain proof; widen checks when dependencies, failures, or unresolved risk warrant it. Relevant changes during inspection invalidate affected conclusions; reassess or report the gap. Match optional methods and verifier edits to active selections. For setup changes, north-star conflicts or missing agreed updates affecting acceptance, assess [project readiness](../../references/project-readiness.md) read-only. When a project verifier supports acceptance or changed in this assignment, read the [verification guidance](../../references/verification-work.md). Compare the verifier, covered user paths, and evidence with the plan and implementation; for commissioned changes, also inspect the accepted proposal and closing trial. Report defects without fixing them. Lead with a reasoned judgment: goal achieved, corrections needed, or open points. Failed necessary checks prevent goal achieved. Missing necessary proof, unclear scope, or missing plan versions remain explicit, even when there are also correctable defects. Keep both axes visible: plan or spec deviation, and repository standards or simplicity. On a simplicity question, use available Efficiency guidance. Investigate `any` or forced casts, repeated workarounds, and unanticipated parameters or state in context. When a change or fix is a band-aid, such as a symptom patch, special case, duplicated logic, or stacked workaround, and a cleaner solution fits the approved plan, report an actionable finding with the evidence and that alternative. A redesign outside that scope stays a recommendation for the human; it does not expand correction or block a positive review. Assess inherited and new learning candidates from the reviewed work. When any exist, follow the learning guidance linked by the working agreement: reconcile the whole collection and offer newly confirmed or materially changed lessons. On a final positive Review, surface an unchanged eligible offer once if it was neither accepted nor rejected; do not restate it on an intermediate correction Review. Keep Review read-only. For each actionable defect, explain observed behavior, evidence, impact on the plan, required correction, permitted scope, and recheck. This can serve as the correction assignment. In standalone work, recommend that the human commission the corrections. In Auto-Work, return findings to the commissioned loop. For missing proof, name the check and a permitted way to obtain it; verification-only correction needs no code change. Consequential decisions or missing access remain explicit. Do not ask the human to attest evidence the executor can collect under its assignment. A positive Review establishes repository acceptance criteria, not Auto-Work's remaining human acceptance or delivery. Finish without starting correction or delivery.
From geldmacher/workflow, MIT.
More from geldmacher/workflow
- 0.603score
verify-auto-work
geldmacher
Verify built Workflow Auto-Work behavior in isolated native-agent trials and local delivery simulations.
SkillAFreshness: A★ 3Claude Code +9 - 0.600score
deploy-local-plugin
geldmacher
Build, inspect, dry-run, or deploy the current repository's host-specific local plugin copies for Cursor, Codex, and OpenCode.
SkillAFreshness: A★ 3Claude Code +9 - 0.598score
verify-project-setup
geldmacher
Verify Workflow project setup, maintenance, readiness offers, and verifier cooperation in isolated native trials.
SkillAFreshness: A★ 3Claude Code +9 - 0.598score
auto-work
geldmacher
Experimental: run requested Light/Dark sequences with independent review.
SkillAFreshness: A★ 3Claude Code +9 - 0.596score
implement-work
geldmacher
Implement a supplied plan artefact outside the host Implement Plan action.
SkillAFreshness: A★ 3Claude Code +9 - 0.595score
setup-workflow
geldmacher
Establish project purpose, checks and agent guidance.
SkillAFreshness: A★ 3Claude Code +9 - 0.595score
release-plugin
geldmacher
Choose the appropriate semantic version and complete this repository's validated Cursor, Codex, and OpenCode GitHub Release in one invocation.
SkillAFreshness: A★ 3Claude Code +9 - 0.595score
plan-work
geldmacher
Plan repository implementation when requested.
SkillAFreshness: A★ 3Claude Code +9 - 0.595score
verification-work
geldmacher
Create or maintain project verifiers.
SkillAFreshness: A★ 3Claude Code +9 - 0.595score
engineering-work
geldmacher
Recommend playbooks; apply human-selected methods.
SkillAFreshness: A★ 3Claude Code +9 - 0.593score
explain-work
geldmacher
Explain Workflow plans, review findings, or handoffs read-only.
SkillAFreshness: A★ 3Claude Code +9 - 0.593score
install-release
geldmacher
Install or update Workflow from its latest stable GitHub Release for Cursor, Codex, or OpenCode.
SkillAFreshness: A★ 3Claude Code +9
Signals
- Stars
- 3
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 21, 2026 (8 days ago)
- Last release
- Oct 21, 2026 (8 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/geldmacher/workflow
- Website
- dennis.geldmacher.dev
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from geldmacher
- 0.603score
verify-auto-work
geldmacher
Verify built Workflow Auto-Work behavior in isolated native-agent trials and local delivery simulations.
SkillAFreshness: A★ 3Claude Code +9 - 0.600score
deploy-local-plugin
geldmacher
Build, inspect, dry-run, or deploy the current repository's host-specific local plugin copies for Cursor, Codex, and OpenCode.
SkillAFreshness: A★ 3Claude Code +9 - 0.598score
verify-project-setup
geldmacher
Verify Workflow project setup, maintenance, readiness offers, and verifier cooperation in isolated native trials.
SkillAFreshness: A★ 3Claude Code +9 - 0.598score
auto-work
geldmacher
Experimental: run requested Light/Dark sequences with independent review.
SkillAFreshness: A★ 3Claude Code +9 - 0.596score
implement-work
geldmacher
Implement a supplied plan artefact outside the host Implement Plan action.
SkillAFreshness: A★ 3Claude Code +9 - 0.595score
release-plugin
geldmacher
Choose the appropriate semantic version and complete this repository's validated Cursor, Codex, and OpenCode GitHub Release in one invocation.
SkillAFreshness: A★ 3Claude Code +9
Cite this page
review-work skill. Agent Mart. https://agentmart.dev/skills/geldmacher-review-work (last updated 2026-10-24)