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 skills add vitaliikapliuk/modelharness --skill verification-loopgit clone --depth 1 https://github.com/vitaliikapliuk/modelharnessWrote 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/vitaliikapliuk/modelharness/verification-loop)<a href="https://agentmods.dev/skills/vitaliikapliuk/modelharness/verification-loop"><img src="https://agentmods.dev/badge/skills/vitaliikapliuk/modelharness/verification-loop.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 5.1 | $0.00038 | $0.00361 |
| Opus 5 | $0.00019 | $0.00180 |
| Sonnet 5 | $0.00008 | $0.00072 |
| Haiku 4.5 | $0.00004 | $0.00036 |
Grade A, and why
verification-loop 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 8d 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.
What it actually says
Verification Loop
Fable 5 builds its own checking harness on long tasks. Reproduce that explicitly.
1. Define "done" before starting
Write down (in your todo list or a scratch note) the binary checks that prove the task is complete: which tests must pass, which command must produce which output, which behavior must be observable. If the user's request doesn't define them, derive them and state them in one sentence.
2. Verify on a cadence, not at the end
After each meaningful increment (a function implemented, a module refactored), run the relevant real check immediately — the actual test command, the actual build. Never batch all verification to the end of a long task; errors compound.
3. Fresh-context verification before completion claims
Before declaring a substantial task complete, dispatch a verifier subagent
(modelharness:verifier agent, or a general-purpose subagent given the
agents/verifier.md instructions). Give it: the original specification, the
definition-of-done checks, and the current diff (git diff output or changed file
list). A fresh context catches what you have rationalized away. Self-review is the
fallback only when subagents are unavailable.
4. Act on the verdict
If the verifier reports gaps: fix them, re-run the checks, re-verify. Only report completion to the user when checks pass AND the verifier (or the checks themselves) confirm it. Report any unverified remainder explicitly.
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.
- 8d ago First seen · 33 lines · 38 tokens per session scan A 735f610d5b77
verification-loop is a skill published in the GitHub repository vitaliikapliuk/modelharness (38 stars, last pushed 2mo ago), licensed MIT. It adds 38 tokens to every session and 361 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
push-ci
Push to remote and monitor CI. Validates branch safety, executes git push WITH explicit user approval, then monitors CI run status via gh CLI. Use when: user says 'push', 'push and watch CI', 'ship it', 'push-ci'. Not for: committing (use /smart-commit), creating PRs (use /create-pr), merging (use /merge-prep).
create-request
Create, update, or scan per-task request tickets for progress tracking. These are date-prefixed non-lifecycle docs under requests/, NOT feature-level requirements (use /req-analyze for those). Use when: tracking task progress, updating completion status, scanning incomplete requests, checking request status dashboard.…
codex-setup
Initialize sd0x-dev-flow infrastructure for Codex CLI and other non-Claude agents. Generates AGENTS.md, installs the commit-msg hook, copies runner scripts. The pre-push gate is opt-in via --with-push-gate. Use when setting up a new project or after updating skills.
smart-rebase
Smart partial rebase for squash-merge repositories. Auto-detect which commits to keep/drop when base branch was squash-merged into target. Use when: user says 'rebase', 'partial rebase', 'base already merged', 'smart rebase', or /smart-rebase. Not for: simple git rebase (the developer runs it — Claude never executes…
remind
A correction helper that checks which project rules or required steps were missed and loads the relevant guidance.
adr
Write an Architecture Decision Record (ADR) for a feature — Context / Decision / Status / Consequences / Alternatives, filed as docs/features/ /adr- - .md with a 3-digit zero-padded number. Handles the Superseded case: bidirectional linking when a new ADR replaces an old one. Use when: recording why an architectural…