# Perspicuity > Perspicuity guides AI assistants to keep a record of delegated work, decisions and evidence. Set the outcome and limits, then review the result. > An open-source skill from People and Planet Consulting Inc. The skill is free. > Your assistant may have its own charges. ## Main pages - [Home](https://perspicuity.ai/): the practice, a fictional worked example, the brief worksheet, and the public toolkit and support links. The site is a single page. Its sections are anchored at `#practice`, `#tools`, `#start`, `#ideas` and `#participate`. Individual tools are anchored at `#tool-skill`, `#tool-eligibility`, `#tool-worksheet`, `#tool-dictate`, `#tool-siteplan`, `#tool-sitewalk` and `#tool-agenttrace`. ## Tools The software and website check are free. Some tools require a paid assistant or API. The Perspicuity skill is always first and the agent eligibility check is always second. The same list is in this page's structured data as a schema.org ItemList at `https://perspicuity.ai/#tools`. - [The Perspicuity skill](https://github.com/perspicuity-ai/decision-toolkit/releases/tag/v0.7.0-preview.3): an open-source preview that carries one piece of work through Frame and Decide, Act and Review in a single record, working in Plan, Run or Review mode. [Installation](https://github.com/perspicuity-ai/decision-toolkit/blob/v0.7.0-preview.3/INSTALL.md). - [Agent eligibility check](https://agents.perspicuity.ai/): enter a domain, get a dated report on its homepage, crawler rules and structured information. 17 checks, each with its evidence. Starter files need business details and review before use. No model decides the result. [Rubric](https://agents.perspicuity.ai/rubric/) · [Decision record](https://agents.perspicuity.ai/corpus/) · API: `POST https://agents.perspicuity.ai/api/audit` with `{"domain":"example.com"}`. - [The brief worksheet](https://perspicuity.ai/#start): three questions that produce a delegation brief for any assistant. Runs in the browser and sends nothing to us. - [dictate](https://github.com/perspicuity-ai/tools/tree/main/tools/dictate): hotkey voice dictation for Linux with X11. Audio goes to an external transcription service using your API key. API charges apply. Published with its measurements and its abandoned approaches. - [siteplan](https://github.com/perspicuity-ai/siteplan): turns a site's intent into an agent-first design brief and a machine-readable plan a built site can be checked against. Every recommendation is labelled `published practice` or `our judgement`; an audit demoted 5 of the 15 practice-labelled claims to judgement, and no project has yet gated a build on one of its plans. [Decision record](https://github.com/perspicuity-ai/siteplan/blob/main/RECORD.md). - [sitewalk](https://github.com/perspicuity-ai/sitewalk): maps a site's structure by crawling it or by reading a built directory, with a `--strict` mode a release step can gate on. On its first live run it reached all 40 URLs the sitemap names and 10 it does not; 4 of its 7 error findings there were its own fault, 57%, and the fix is verified on a reproduction rather than a live re-run. It does not execute JavaScript. [Decision record](https://github.com/perspicuity-ai/sitewalk/blob/main/RECORD.md). - [agenttrace](https://github.com/perspicuity-ai/agenttrace): reads a web server's access log and reports what claimed AI agents asked for, including whether `/llms.txt` is read. A user-agent string is self-declared and spoofable, so every count is a count of claims, and the tool says so in its own output. It keeps a site's own monitor out of that count; on its first real log it found no external AI agent traffic in a 510-second window. [Decision record](https://github.com/perspicuity-ai/agenttrace/blob/main/RECORD.md). ## Machine-readable surfaces - [Sitemap](https://perspicuity.ai/sitemap.xml) - JSON-LD in the home page states the organisation, the service, the preview and the tool list. ## What this site does not claim The preview is an early project. Comparative benefits for delegated work, comprehension and any emissions saving remain unmeasured, and the site states its evidence limits where it makes a claim. ## Reuse The public toolkit and this site's source are published under their stated licences. Contact `david@peopleandplanet.consulting` before quoting a claim as a result.