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 PromptPartner/agentsmith --skill verifygit clone --depth 1 https://github.com/PromptPartner/agentsmithWrote 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/promptpartner/agentsmith/verify)<a href="https://agentmods.dev/skills/promptpartner/agentsmith/verify"><img src="https://agentmods.dev/badge/skills/promptpartner/agentsmith/verify/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.
<a href="https://agentmods.dev/skills/promptpartner/agentsmith/verify"><img src="https://agentmods.dev/badge/skills/promptpartner/agentsmith/verify.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00069 | $0.00697 |
| Opus 5.5 | $0.00028 | $0.00279 |
| Sonnet 5.5 | $0.00014 | $0.00139 |
| Haiku 4.5 | $0.00007 | $0.00070 |
Grade A, and why
verify 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 1mo 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 — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verify — evidence before "done"
"Verified" is not a feeling; it's output you can point at. Never claim passing without showing it (R5 + verification-before-completion).
When this fires
"verify" / "is this done / shippable / ready to merge?" / just before a commit or PR.
Runtime neutrality
The verification runner and evidence standard are client-neutral. Never infer the active agent
from this skill's install path: portable canonical content may be copied into a required client
adapter such as .claude/skills. Use canonical AGENTS.md when an instruction is relevant.
Fast path — if the Agentsmith CLI is available
- Run
agentsmith verify(or the installed-project shim.agentsmith/agentsmith verifyon macOS/Linux and.agentsmith\\agentsmith.cmd verifyon Windows). It runs every phase in.harness/verify.confand stops at the first failure).--listshows the phases;--only <tag>iterates just one. Add--record <directory> --tree-class <clean-clone|disposable-fixture|linked-worktree|operator-worktree>when deterministic command evidence needs a durable local receipt. The explicit class keeps a dirty operator tree from masquerading as clean-checkout evidence;clean-clonerefuses a dirty tree. The destination must be new; relative paths resolve from the project target. Record mode redacts secret-shaped console and sidecar output before either is emitted. - On a failure: read the label + command it printed, explain in plain language what broke, and
point at that phase's line in
.harness/verify.confto fix or refine. - Report the actual pass/fail output — not a summary of intent.
The v1 receipt covers deterministic commands only. Keep screenshots, videos, manual assertions, published results, and real runtime or visual checks as separately referenced artifacts. A green receipt does not satisfy an end-to-end gate that requires those observations.
Fallback — no runner or no conf
- Say so plainly, and look at
.harness/verify.conf.examplefor the intended phases. - Run the obvious checks for this project directly and show their output: build, test, lint, link/render check, a dry-run. For non-code work, "verify" = open the artifact and confirm it renders / the numbers reconcile / the links resolve.
- If nothing is wired yet, propose the phases this project needs and offer to write
.harness/verify.conf— but still run the checks by hand this time.
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.
- 1mo ago Changed · +2 lines a40042bce40d
- 1mo ago Changed · +7 lines 44676e7d6fc9
- 1mo ago First seen · 40 lines · 69 tokens per session scan A e5431963c657
verify is a skill published in the GitHub repository PromptPartner/agentsmith (178 stars, last pushed 7d ago), licensed MIT. It adds 69 tokens to every session and 697 once invoked, about $0.0003 per session on Opus 5.5. 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
meta-reviewing-ai-reviewing
AI integration review patterns. Use when reviewing model API calls, prompt construction, LLM output handling, RAG pipelines, and tool-calling code. Covers prompt-injection call-chain tracing, output validation, stop reasons, token budgets, retry and timeout handling, streaming, and key/PII exposure.
meta-reviewing-api-reviewing
Backend code review — injection and SSRF, boundary validation, object-level authorization, mass assignment and data exposure, error leakage, N+1 and unbounded queries. Use when a diff touches API routes or handlers, database access, auth middleware, or server code that reads files, starts processes or fetches URLs.
meta-reviewing-cli-reviewing
CLI code review — exit codes, cancellation and signals, non-TTY behaviour, the stdout/stderr split, error text, config precedence, subprocess safety and test adequacy. Load when reviewing a command-line tool's commands, entry point, prompts, output or tests, whichever parser or prompt library it uses.
meta-reviewing-infra-reviewing
Infrastructure code review — supply-chain pinning, secret exposure, workflow triggers and untrusted input, container hygiene, deployment safety. Load when reviewing a diff that touches CI/CD workflows, Dockerfiles, deployment configs or IaC.
meta-reviewing-web-reviewing
React component review — hook rules, effect cleanup, derived state, keys, memoization evidence and keyboard access. Load when reviewing a diff that adds or changes React components or hooks.
meta-reviewing-reviewing
Code review process — reading order, findings, severity, scope and the approval decision. Load when reviewing a diff or PR, writing review comments, or deciding whether to approve or request changes.