run-harness

run-harness is a skill for Claude Code from Viindoo/odoo-mcp-client. It costs 125 tokens per session (7,855 once invoked), scanned A, original, MIT.

A workflow runner that carries out an already-approved multi-step plan until it finishes or reaches a step needing human approval. It passes each step to the specialist guide responsible for that work and records the result.

In plain words
What is it for?
Use it to drive serialized multi-step work, dispatch specialist tasks in order, track outcomes, and continue a run through its dependency graph.
Why use it?
It removes the need to manually prompt the agent after every completed step. It also keeps the shared run state and stops when required context or human approval is missing.

Skill for Claude Code

Written for Claude Code: user-invocable in frontmatter. Also seen: model in frontmatter; mentions subagents.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the odoo-ai-agents plugin — 51 skills, 8 commands, 26 agents, 8 hooks, 1 MCP server shipped together

Good fit Use it to drive serialized multi-step work, dispatch specialist tasks in order, track outcomes, and continue a run through its dependency graph.

Compare 6 skills from other repositories ↓
Install

Getting it into your agent

This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.

Claude Code
/plugin marketplace add Viindoo/odoo-mcp-client
Claude Code
/plugin install odoo-ai-agents

Made for: Claude Code.

Or install odoo-ai-agents, the plugin that ships this one along with the rest of its 51 skills, 8 commands, 26 agents, 8 hooks, 1 MCP server.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for run-harness

README.md
[![agentmods](https://agentmods.dev/badge/skills/viindoo/odoo-mcp-client/run-harness/github.svg)](https://agentmods.dev/skills/viindoo/odoo-mcp-client/run-harness)
Your own site
<a href="https://agentmods.dev/skills/viindoo/odoo-mcp-client/run-harness"><img src="https://agentmods.dev/badge/skills/viindoo/odoo-mcp-client/run-harness/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for run-harness

Your own site · 80×15
<a href="https://agentmods.dev/skills/viindoo/odoo-mcp-client/run-harness"><img src="https://agentmods.dev/badge/skills/viindoo/odoo-mcp-client/run-harness.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 125 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 7,855 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00125 $0.07855
Opus 5 $0.00063 $0.03927
Sonnet 5 $0.00025 $0.01571
Haiku 4.5 $0.00013 $0.00785

Measured 6d ago against content hash 402143cfbf34, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

run-harness scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/odoo-ai-agents/skills/run-harness/SKILL.md · 451 lines

How it starts

The opening of the file, as written. The whole thing — 451 lines — stays where its author put it; the contents beside it link to each section on GitHub.

run-harness - Drive-to-done loop

Role

Conductor of a multi-step run. Owns no domain expertise: reads the blackboard, decides the next step, dispatches it, records the result. Prompt-discipline plus advisory hook nudges - NOT a hard scheduler (never trap the main agent). Mechanism SSOT: docs/reference/workflow-harness.md §8 - this file is the operating procedure, that is the contract.

Out of Scope

  • Authoring business artifacts → dispatches the specialist; only writes run-<id>.json.
  • Planning / serializing the DAG → intake's Phase P. The driver only walks an EXISTING run-<id>.json; it never ingests a plan .md. A plan-producing skill (e.g. odoo-planning) routes it to intake Phase P (next: odoo-intake), NEVER next: run-harness with only a plan pointer - reaching this loop before serialization yields NEEDS_CONTEXT.
  • Coercing the main agent → advisory nudges only (Hard rule #2).
  • Crossing the Odoo↔general boundary → intake's routing decision.

Hard rules

  1. Owns the blackboard. This is the orchestrator that walks the RUN-DAG; it owns the run state and controls dispatch, so it MUST NOT be invoked from inside a subagent.
  2. Never hard-block the main agent - but that is not an open license to pause. This loop is prompt-discipline, not coercion: the Stop/PreToolUse hooks only nudge (advisory); they never deny a tool call or block a turn-end (quality-gate block is only ever for a subagent). That does NOT make "a node just finished" a reason to stop. The main agent auto-advances the run WITHOUT asking a human UNLESS one of the following ENUMERATED conditions holds - these are the ONLY legitimate reasons to end the turn and await a human "continue": (a) the node is L2 (§ Gate-tier resolution below - ALWAYS a human gate); (b) the run resolves to BLOCKED for ANY reason (§ Circuit-breakers below and every "STOP BLOCKED" in this file - examples, not a closed set); (c) NEEDS_CONTEXT; (d) the human issued an explicit stop/abort phrase (§ Circuit-breakers). Finishing a node, or any single subagent dispatch, is by itself never one of these - and a run that plows past a genuine (a)-(d) condition instead of stopping is BLOCKED behavior too (a real blocker ignored), not drive-to-done. (a)-(d) are the reasons to end a turn AWAITING A HUMAN; ending one to RECEIVE a dispatched agent's report is neither - that stop needs no human: the launch is asynchronous, so stopping is the ONLY point a result can reach you (§ The loop > dispatch), and the run resumes on the child's own completion.
  3. Writes to run-<id>.json after creation are run-harness's alone. Intake's Phase P performs the one-time bootstrap write; from then only this loop writes it (hooks never write it - no race).
  4. You dispatch; subagents do not. A step emits a Continuation Contract (a signal); acting on its next[] is THIS loop's job. Respect the worker-brief contract (snippets/worker-brief.md).
  5. L2 is always a human gate. The autonomy dial can lower L1→auto-pass but NEVER L2 (irreversible/outward: shared instance, git MERGE to the principal branch, send to a third party). The merge is the ONLY OUTWARD L2 (owned by odoo-pr-monitoring) - not the only L2 a run hits: the REGISTRY also returns L2 for an instance-touching skill (odoo-i18n, odoo-acceptance), and the ephemeral ceiling lowers exactly the nodes THIS driver briefed against a throwaway database. No local merge into the principal checkout; no auto-merge. EVERY node's tier, integrate included, comes from the ONE total function in § Gate-tier resolution, never from a prose exception.
  6. Worktree-always for SOURCE-writing dispatch (realizes intake Hard Rule 6). Before dispatching a node that writes the SOURCE tree (not the $ODOO_AI_HOME state root; same test as Gate-tier resolution), if it has no WORKTREE_PATH/TARGET: worktree:<path> and its approach is not a self-provisioning specialist (SSOT list: ${CLAUDE_PLUGIN_ROOT}/snippets/git-delegation.md § Self-provisioning specialists), INVOKE git-toolkit:git-ops for a dedicated worktree/branch, inject its path into the node's inputs, and write(RUN). NEVER dispatch a source-writing node against the principal checkout. A node writing only under the $ODOO_AI_HOME state root (e.g. odoo-code-review at TARGET=local) is NOT provisioned.

Read the full file on GitHub · 451 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 6d ago First seen · 451 lines · 125 tokens per session scan A 402143cfbf34

Subscribe to this mod's changes

run-harness is a skill published in the GitHub repository Viindoo/odoo-mcp-client (7 stars, last pushed 3d ago), licensed MIT. It adds 125 tokens to every session and 7,855 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

local-ai-agents

Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…

microsoft/ai-agents-for-beginners · 200 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

insight-error-page

Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…

vercel/next.js · 83 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens

next-partial-prefetching-adoption

Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…

vercel/next.js · 103 tokens