From agent-verify
Use before a final response when Codex has made or is about to make claims about tests passing, files changed, commits, pushes, or pull requests. Runs Verify in notification mode and reports what was false, missing, or inconclusive.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-verify:verify-claimsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill before finalizing an answer that claims work was completed.
Use this skill before finalizing an answer that claims work was completed.
Verify is a notification workflow in Codex. It does not have Claude Code's Stop-hook enforcement, so you must run the notifier and include any failures or inconclusive checks in the final response.
Run Verify when the final answer includes claims like:
From the repository root, run:
npm run codex:notify -- --message "<final answer draft>"
On Windows shells where npm is blocked, use npm.cmd:
npm.cmd run codex:notify -- --message "<final answer draft>"
If the user asks to verify a different project, pass its path:
npm run codex:notify -- --cwd "/path/to/project" --message "<final answer draft>"
If Verify prints Verify notification: all detected claims checked out., you may answer normally.
If Verify reports Not done or unverified, include those items in the final answer and correct any success claims. Do not say tests passed, files changed, commits were made, branches were pushed, PRs were opened, or protected sections stayed intact when Verify says they failed or were inconclusive.
If Verify reports No supported completion claims detected., no verification note is needed.
CLAUDE_TEST_HANDOFF.md while using this skill.npx claudepluginhub orthogon-ai-labs/agent-verify --plugin agent-verifyProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.