Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/igovet/codex-cortex-orchestrator/output-validationnpx skills add igovet/codex-cortex-orchestrator --skill output-validationgit clone --depth 1 https://github.com/igovet/codex-cortex-orchestratorWrote 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.
[](https://agentmods.dev/skills/igovet/codex-cortex-orchestrator/output-validation)<a href="https://agentmods.dev/skills/igovet/codex-cortex-orchestrator/output-validation"><img src="https://agentmods.dev/badge/skills/igovet/codex-cortex-orchestrator/output-validation.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00029 | $0.00984 |
| Opus 5 | $0.00015 | $0.00492 |
| Sonnet 5 | $0.00006 | $0.00197 |
| Haiku 4.5 | $0.00003 | $0.00098 |
Grade A, and why
output-validation 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 4d 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.
How it starts
The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Output Validation
Validate the result, not the confidence of its author. The coordinator designs the evidence request and delegates project verification; a validation worker performs every project inspection and executed check. The coordinator never reads project artifacts, runs a command, or performs substantive verification itself.
- Restate acceptance criteria as observable checks and create bounded validation delegations. Compose the orchestrator's authoritative delegation knowledge contract into each worker's instructions.
- Have workers inspect the relevant change and execution path for scope and unintended effects.
- Have workers run the smallest relevant build, typecheck, lint, test, reproduction, or behavioral check; use an independent worker where risk justifies it.
- Reconcile reported files, commands, outputs, and observed behavior by report ID. If the evidence is insufficient, delegate the missing check.
- From worker reports, classify each criterion as passed, failed, blocked, or unverified, with its evidence, risk, and useful next owner.
- Preserve a concise verified change-surface summary for the coordinator's later documentation-impact decision, covering behavior, architecture, interfaces, commands, verification, conventions, and feature ownership.
Build the validation matrix from the original task contract, not from the latest implementation summary. Every independent effective-contract outcome must appear exactly once with its current disposition and canonical report digest. Keep its linked acceptance and verification facts in that single matrix row rather than creating a second row. Before the first publication attempt, preserve the complete ordered server-owned reconciliation row set and compare the finished row count and references against its receipt. Compare the matrix across discovery, every plan revision, implementation assignments, rework, and final verification. An omitted item, a merged finding, or an item present only in an earlier report is a failed handoff and requires rework.
Artifact verification includes existence, absence, unchanged-state, manifests,
Git, caches, worktrees, and project-local .codex. A validation worker performs
every such check; the coordinator never performs one directly.
When validation finds a defect or material concern, use a structured advisory finding in report content: stable finding key, severity, affected surface, sanitized evidence, failure scenario/impact, recommendation, disposition, and any exact decision reference. Preserve conflicting findings and supersede them explicitly; do not turn severity, waiver, or resolution into a backend gate.
For an executed check, the worker records the command, working directory, integer exit code, and decisive sanitized evidence in a report. A passing check has exit code 0. A negative-path check uses an assertion harness that observes the expected failure and itself exits 0. Never omit or relabel a nonzero result.
Missing evidence is not a pass. It may justify more verification, bounded
rework, ready_with_risks, or not_ready; it never becomes a backend
permission barrier or prevents an honest final answer.
Before a material classification, consume the complete canonical report body through the coordinator evidence-read path and finish its continuation chain. A native handoff summary, report reference, projection link, or manifest alone does not establish that findings and unresolved items were transferred.
Record decisive check evidence in the immutable report, including selected large-report sections when chunking is required. A status or final handoff may summarize that evidence only after the relevant report sections were read. When a current contained digest-verified report or plan Markdown projection is available, publish its clickable absolute path with a localized result summary and implication. Task/status/decision/delegation and other non-report records remain SQLite-only. A missing or stale projection produces no link and does not change the acceptance classification.
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.
- 4d ago First seen · 87 lines · 29 tokens per session scan A 058c0e834661
output-validation is a skill published in the GitHub repository igovet/codex-cortex-orchestrator (1 stars, last pushed 4d ago), licensed MIT. It adds 29 tokens to every session and 984 once invoked, about $0.0001 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 2026-08-31.
Other skills, from other repositories
bug-triage
Triage bugs reported in chat/issues, search for duplicates, file or update GitHub issues with full context, and push fix PRs.
ao-desktop-dev
Launch, restart, or troubleshoot the real AO Electron desktop app from this repository; run a checkout against isolated or real local AO data; combine PR branches for local UI review; and diagnose stale Electron processes, port conflicts, or preload bridge mismatches. Use whenever asked to run, open, show, or visually…
using-ao
Catalog of the AO (Agent Orchestrator) ao CLI: spawning workers, managing sessions and projects, sending messages, controlling the shared browser, previewing pages, and daemon control. Use when using the ao CLI, spawning workers, or managing AO sessions in an AO workspace.
summon
Autonomous orchestrator for manifest work items through the development lifecycle. Use when running egregore to process a backlog automatically.
model-router
当需要调度不同模型执行子任务时使用。根据任务类型匹配模型能力,自动选择最优模型并生成调用指令。TRIGGER: 模型调度、model dispatch、选模型、分派任务、多模型协作、路由到模型.
goal-flight
Portable Goal Flight workflow for long-running repo work: planning, dispatch, review, recovery, file-backed resume.