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 fledgeling-co/fledgeling-plugins --skill proctorgit clone --depth 1 https://github.com/fledgeling-co/fledgeling-pluginsWrote 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/fledgeling-co/fledgeling-plugins/proctor)<a href="https://agentmods.dev/skills/fledgeling-co/fledgeling-plugins/proctor"><img src="https://agentmods.dev/badge/skills/fledgeling-co/fledgeling-plugins/proctor/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/fledgeling-co/fledgeling-plugins/proctor"><img src="https://agentmods.dev/badge/skills/fledgeling-co/fledgeling-plugins/proctor.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00450 | $0.14134 |
| Opus 5.5 | $0.00180 | $0.05654 |
| Sonnet 5.5 | $0.00090 | $0.02827 |
| Haiku 4.5 | $0.00045 | $0.01413 |
Grade A, and why
proctor 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 — 963 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Proctor
You are running a test campaign against a native macOS application, using instruments rather than impressions. The distinction runs through everything below: a screenshot you looked at is an impression, a screenshot with a frame status and a dirty-rect summary is an instrument reading. Only one of them can be wrong in a way you would notice.
The campaign produces a report a person can act on, in which every claim is traceable to a tool result, and everything that could not be established is named rather than quietly omitted.
Running as a Gemini model? Read gemini.md in this directory first, then follow this file with the overrides it names. Turns proctor's instrument-reading discipline into a mechanical contract on Gemini: every claim quotes the MCP tool return and field it rests on (there is no exit code to paste), the case matrix, state matrix, capture set and fidelity ledger become filled tables with denominators before the first proctor_act, every stated bound is read back off the delivered campaign, a skipped hand-off leaves its rows open rather than passed, and platform conformance numbers are read from the rubric files rather than recalled. Other models skip it.
What Proctor observes, and why that is the whole point
Proctor can delegate the clicking. It does not delegate the looking, and that asymmetry is the reason a Proctor result is worth more than the same result from anything else that drives a Mac.
Three observation channels are Proctor's own and stay Proctor's own:
- Capture, with frame trustworthiness attached. Apple defines six
SCFrameStatusvalues and makes checking them a precondition of trusting a frame. Everyproctor_capturereportsstatus,contentRect,dirtyRectCount,dirtyArea,framesWaited,trustworthyand acaveatnaming the cause when trust fails. A stale frame is pixel-identical to a correct one, so those fields are the only thing separating them. General screenshot tools return the image and nothing else, which means a caller cannot tell the two apart at all. - The accessibility walk, with a
TreeProvenancesaying how the tree was obtained, whetherAXManualAccessibilitywas applied and where the walk was truncated. - The verdict layer — assertions, the tri-observer
agreecheck, settle reporting, determinism scoring. These are judgements about the application under test, which is a different job from scoring an agent.
What ships with it
4 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.
- 1mo ago Changed 7014538ab2ec
- 1mo ago First seen · 963 lines · 450 tokens per session scan A 44d359a6fcab
proctor is a skill published in the GitHub repository fledgeling-co/fledgeling-plugins (2 stars, last pushed 4d ago), licensed MIT. It adds 450 tokens to every session and 14,134 once invoked, about $0.0018 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 2026-09-04.
Other skills, from other repositories
android_ui_verification
Automated end-to-end UI testing and verification on an Android Emulator using ADB.
solopi-ai
A command-line framework for testing Android apps and devices with SoloPi, including on-device or cloud AI decision models. It manages devices, test cases, recorded interactions, replays, performance history, and evidence.
ios-simulator
Verify and debug native, React Native, Expo, or Flutter apps on an iOS Simulator with agent-device. Use when an agent needs to launch an app, inspect its live UI, tap, type, scroll, validate a code change, collect failure evidence, or reproduce a workflow on an iPhone or iPad Simulator.
dogfood
Systematically explore and test a mobile app on iOS/Android with agent-device to find bugs, UX issues, and other problems. Use when asked to dogfood, QA, exploratory test, find issues, bug hunt, or test this app on mobile.
eas-simulator
Run and control a user's app on a remote iOS/Android simulator hosted on EAS cloud. Read before running any eas simulator: commands - it has the current syntax for this experimental API. Use whenever the user needs a simulator they can't run locally - 'run my app on a cloud simulator', 'use eas simulator to…
flutter-add-widget-test
Implement a component-level test using WidgetTester to verify UI rendering and user interactions (tapping, scrolling, entering text). Use when validating that a specific widget displays correct data and responds to events as expected.