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/mthines/agent-skills/verify-behaviornpx skills add mthines/agent-skills --skill verify-behaviorgit clone --depth 1 https://github.com/mthines/agent-skillsWrote 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/mthines/agent-skills/verify-behavior)<a href="https://agentmods.dev/skills/mthines/agent-skills/verify-behavior"><img src="https://agentmods.dev/badge/skills/mthines/agent-skills/verify-behavior.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 | $0.00218 | $0.02507 |
| Opus 5 | $0.00109 | $0.01254 |
| Sonnet 5 | $0.00044 | $0.00501 |
| Haiku 4.5 | $0.00022 | $0.00251 |
Grade C, and why
verify-behavior scanned grade C with 2 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 5d 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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
3. **Isolation is not optional for Tier 3.** A throwaway worktree, no tracked-file mutation, scratch cleanup, no network by default, never `curl | sh` (`rules/isolation-safety.md`). Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
3. **Isolation is not optional for Tier 3.** A throwaway worktree, no tracked-file mutation, scratch cleanup, no network by default, never `curl | sh` (`rules/isolation-safety.md`). How it starts
The opening of the file, as written. The whole thing — 176 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verify Behavior
Given a behavioral claim about code, or a change that was just applied, decide the cheapest way to get executed proof, run it in isolation, and report the raw result as a receipt.
This skill is the execution engine six-plus call sites in this repo used to hand-roll independently: "detect the toolchain, run something, read pass or fail." It replaces the ad hoc version in each of those with one shared ladder.
This
SKILL.mdis a thin index. Detailed rules live inrules/*.mdand load on demand.
The execute-not-score boundary
This skill does not score — it never assigns a confidence score and never grades pass/fail against an intent.
It runs a command, captures the raw output, and classifies the result against the claim itself as confirms / contradicts / ambiguous / null.
confidence(code)owns the number — this skill supplies sharper evidence to that gate, it does not replace it.- A calling agent's own grading (e.g.
bug-fix-verifier'sFAIL_TO_PASS, theaw-executorPhase 4expectcomparison) stays with the caller — this skill supplies the run-and-observe mechanic underneath that grading, not the grading itself.
See rules/receipt.md for the full contract, including the hard invariant that a null or non-reproducing result drops or contradicts a finding and is never confirmation.
The two consumer shapes
| Shape | Question it answers | Consumers |
|---|---|---|
| Claim-verification | "Is this specific behavioral assertion true?" — read-only, feeds confidence(code) as Evidence |
agents/shared/rules/verification-receipt.md (pr-reviewer Tier 2/3) |
| Change-verification | "Did this applied change produce the expected green/red result?" — a post-apply gate | bug-fix-verifier, feature-pr-verifier, aw-executor Phase 4 checks loop |
Both shapes share the same core: toolchain discovery, isolated execution, and the receipt format.
Only the output framing differs — see rules/receipt.md.
What ships with it
5 files 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.
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.
- 5d ago First seen · 176 lines · 218 tokens per session scan C 9509f622e7cb
verify-behavior is a skill published in the GitHub repository mthines/agent-skills (12 stars, last pushed today), licensed MIT. It adds 218 tokens to every session and 2,507 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ego-browser
Skill "ego-browser" from citrolabs/ego-lite, covering ego-browser, quick start, common helpers, task spaces and control handoff.
temporal-developer
Develop, debug, and manage Temporal applications across Python, TypeScript, Go, Java, .NET, Ruby, and Rust. Use when the user is building workflows, activities, or workers with a Temporal SDK, debugging issues like non-determinism errors, stuck workflows, or activity retries, using Temporal CLI, Temporal Server, or…
x-algorithm
Write X (Twitter) posts that the For You algorithm actually rewards. Grounded in the open-sourced X recommendation system — the Grok-based transformer ranker, Phoenix retrieval, Thunder in-network store, and Grox content-understanding pipeline. Use when the user wants to write a post, thread, reply, or quote; plan a…
opencode-memory
Browse local OpenCode history: sessions, messages, plans, prompt history, and prior decisions. Use when the user says history, previous session, last time, remember, recall, plans, prior work, or when resuming/debugging repeated work where earlier OpenCode context may help. Do not use for fresh tasks or when current…
creating-explainers
Use when creating an interactive explainer - a single self-contained HTML page with hand-built Canvas figures. Handles source-file explainers, topic-driven research explainers, and mixed intake where files provide the spine and research adds support. Trigger phrases include "make an explainer", "turn this paper into…
custom-icons
Create or refine custom icon assets as native or traced SVGs and transparent PNG/WebP files. Use when the user asks for a bespoke icon or cohesive icon set, wants an image traced into a clean vector, or needs a detailed or 3D icon with transparency.