القائمة
House style for runner dry-run notes — tone, section order, and formatting rules for the plan-work-review artifact. Use when writing or revising a runner dry-run note.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Review a vernier loop definition for soundness before promoting or running it — contract coverage, effect-scope minimality, provider-agnostic role naming, policy termination, retry-feedback threading, trust level honesty, and skill wiring. Use when reviewing a new or changed loop, auditing loops in a repo, or deciding whether a loop is ready to run live.
Operate and debug vernier runs — run/tick/resume semantics, leases, reading journal.jsonl entry types, evidence bundles, show/stats roll-ups, doctor diagnostics, exit codes, and structured JSON errors. Use when running loops, diagnosing a failed or stuck run, resuming after a crash, reading a run ledger, or analyzing usage and cost across runs.
Set up vernier (the agent-orchestration kernel) in a new or existing project — install, create vernier.config, choose where run/eval data is stored (the ledger root), scaffold a starter loop, and verify the machine with doctor. Use when adding vernier to a project, deciding where journals and evidence live, or troubleshooting "no loops registered" / config discovery.
Author Agent Skills (agentskills.io format) for vernier steps — SKILL.md frontmatter rules, registration and discovery tiers, binding skills to steps via the executor-mirroring chain, native vs prompt delivery, and the symlink-free constraint. Use when creating a skill for a loop step, registering or binding skills in vernier.config, or debugging why a skill is not resolving or not being delivered.
Author vernier loops — the five-slot Loop model (Signature, Steps, Policy, Trust, Ledger), Step anatomy (zod signatures, executors as role ids, contracts, effect scopes, prompts, skills), policy combinators, and out-of-tree registration via vernier.config. Use when creating a new loop, adding a step, writing a contract or policy, or converting a workflow into loop-as-data.
House greeting style. Use when greeting a reviewer in any generated text.
| name | dry-run-note-style |
| description | House style for runner dry-run notes — tone, section order, and formatting rules for the plan-work-review artifact. Use when writing or revising a runner dry-run note. |
Write the note so a reviewer can scan it in under a minute.
Runner dry run <trace-id> — the trace id verbatim, nothing else.## Route, ## Bundle,
## Runner Verification, ## Improvement Candidate.label: value lines over prose.## Route: who approved (the route gate), the worker role, and the
one-sentence routing reason.## Bundle: the bundle path and the artifact path, each on its own
labeled line.## Runner Verification: state plainly that the runner validates this
artifact and writes the trace — the worker does neither.## Improvement Candidate: exactly one candidate, phrased as an
imperative sentence naming a concrete loop change, not a vague wish.