Review gate
Reusable, token-optimized agent skills for autonomous development with Hermes, Claude Code & OpenCode. Multi-model orchestration patterns — context-checkpointing, human review gates, RLS security, GDPR consent, AI cost control, and test-fix loops. Framework-agnostic, agentskills.io compatible.
npx -y skills add prasadmogulothu/agent-skills --skill review-gateAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when a task is marked as needing human review. Pause, present the artifact to a human (summary + link), and wait for explicit approval before continuing. Loop on change requests; never proceed on silence.
SKILL.md
1.7 KB, as published. Nobody here has run it
review-gate
Human-in-the-loop approval as a reusable step. Mark certain tasks as gated; at those, a human must approve before dependent work continues.
Steps
- Confirm the task passes your done-checks.
- Set the task
Status: BLOCKED_REVIEWintasks.mdand add it to a Review Queue. - Notify the human (chat / messaging gateway / PR comment) with:
- What — a 3-bullet summary of what was built or changed.
- Where — a clickable artifact: a git diff/PR, a preview URL, a component catalogue, or a report.
- Ask — "Approve, or request changes?"
- Wait. Do not start any task that depends on this one. You may pick up an independent, non-dependent task meanwhile to keep momentum.
- Never proceed on silence. Only an explicit approval advances the gate.
Change loop
On a change request: set Status: CHANGES_REQUESTED, record the request, apply the fix with the
same-tier model, re-run done-checks, re-present, and ask again. Loop until approved, then set
DONE and clear it from the Review Queue.
Recommended always-gated work
Anything risky or irreversible: data-model/security changes, auth, payments, data deletion, and public deploys — gate these even when automated checks pass.
Token tactics
Near-zero model cost — runs in the agent's control flow. Because it pauses the workers while a human reviews, models sit idle (no spend) until approval. Oversight that also saves tokens.