[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-83160":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":17,"stars90d":16,"forks30d":16,"starsTrendScore":18,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":16,"starSnapshotCount":16,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},83160,"agentic-readiness-guide","forter\u002Fagentic-readiness-guide","forter","Agentic Readiness Guide","",null,"JavaScript",99,4,63,1,0,36,7,2.1,"Creative Commons Attribution 4.0 International",false,"main",[24],"team-field-cto-office","2026-06-12 02:04:31","# Forter Agentic Readiness Guide\n\nA practical, opinionated guide to making any website **agent-ready** - discoverable, comprehensible, trustworthy, actionable, and experiential - so that LLM-driven agents (and the humans behind them) can find, understand, trust, and act on your product.\n\nFive modules, 25 guidelines, each one a single markdown file in [`content\u002F`](.\u002Fcontent). Every guideline has a \"What & why\", a 1-5 complexity\u002Fimpact score, concrete steps, and references. Where Forter ships infrastructure that satisfies a guideline, the file ends with a \"How Forter helps\" callout.\n\n## Download it offline\n\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F9817a638-6ffc-42b3-94b4-a0124f280cea\n\n**[Download the full guide (PDF)](https:\u002F\u002Fl.forter.com\u002Fhubfs\u002FForter-agentic-readiness-guide.pdf)**\n\n## Read it online\n\nBrowse [`content\u002F`](.\u002Fcontent) directly on GitHub. Files are organized by module: `m1-*` Discoverable, `m2-*` Comprehensible, `m3-*` Trustworthy, `m4-*` Actionable, `m5-*` Experiential.\n\n## Audit your own site\n\nThis repo ships a Claude Code skill in [`SKILL.md`](.\u002FSKILL.md), backed by 25 machine-testable rubrics in [`audit\u002F`](.\u002Faudit), that turns the guide into an automated auditor. Point it at a site (and optionally its source) and it will:\n\n1. Run the probe in each `audit\u002Fm{M}-{N}.md` rubric against your site.\n2. Score each guideline **Pass \u002F Partial \u002F Fail \u002F N\u002FA**, citing the literal probe response as evidence.\n3. Rank Fails and Partials by **impact × (6 - complexity)** so the highest-leverage fixes float to the top.\n4. Optionally apply fixes as commits when you give it your repo path.\n\n### Install the skill\n\n[Claude Code](https:\u002F\u002Fdocs.claude.com\u002Fclaude-code) auto-discovers skills under `~\u002F.claude\u002Fskills\u002F`. Clone once and **symlink** the repo in - no copying, so `git pull` keeps the skill current and your skills folder stays clean:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fforter\u002Fagentic-readiness-guide.git\nmkdir -p ~\u002F.claude\u002Fskills\nln -s \"$(pwd)\u002Fagentic-readiness-guide\" ~\u002F.claude\u002Fskills\u002Fforter-agentic-readiness-audit\n```\n\n(The symlink's name matches the skill's `name`. The repo's root `SKILL.md` and `audit\u002F` resolve straight through the symlink. To uninstall, `rm ~\u002F.claude\u002Fskills\u002Fforter-agentic-readiness-audit` - that removes only the link, not your clone.)\n\nThen in any Claude Code session:\n\n```bash\nclaude \"Audit https:\u002F\u002Fyour-site.example.com against the Agentic Readiness Guide\"\n```\n\nClaude picks the skill up from its frontmatter `description` and runs it. Add `--add-dir \u002Fpath\u002Fto\u002Fyour\u002Fsite` to include your source repo - fixes get applied as commits there. Type `\u002Fskills` inside Claude Code to confirm the skill is loaded.\n\n**Prereqs.** The probes shell out to `curl`, `jq`, and `python3` (for HTML parsing). All three are standard on macOS and most Linux distros; on a bare container, install via `apt-get install curl jq python3` or `brew install jq` (curl and python3 usually ship).\n\n### What you get\n\nA single `report\u002FAUDIT.md` with, in order:\n\n- **One-line scoreboard** - `Score N\u002FW (P%) · X Pass · Y Partial · Z Fail · K N\u002FA`.\n- **Headline** - two sentences a non-technical reader can grasp: where the site sits and the rough shape of the gap.\n- **Action plan** - ordered fixes (smallest concrete change → guideline it unblocks → effort → point gain), with a projected post-fix score.\n- **Cross-cutting blockers** - when one bug gates ≥ 3 guidelines, it gets a dedicated reproduce-and-fix subsection instead of being repeated per row.\n- **Findings table** - 25 rows, one per guideline, with the literal probe response as inline evidence.\n\nRaw probe outputs land alongside in `report\u002F*.out`; opt into `report\u002Fscore.json` (machine-readable, schema below) and a PR-ready issue list by asking for them. A typical first audit on a real e-commerce site closes the discoverability tier in a day and surfaces 15-20 deeper items the team can sequence over a sprint.\n\n### `report\u002Fscore.json` (machine-readable, opt-in)\n\nAsk for it explicitly (\"also write `score.json`\") and the skill emits a single JSON file next to `AUDIT.md` - the same scores as the report, structured for CI\u002FCD gates, dashboards, or trend tracking. The findings table is for humans; `score.json` is for machines.\n\n```jsonc\n{\n  \"host\": \"example.com\",                 \u002F\u002F bare host audited (no scheme)\n  \"tested_at\": \"2026-06-02T14:30:00Z\",   \u002F\u002F ISO-8601 UTC timestamp of the run\n  \"scope\": [\"m1-1\", \"m1-2\", \"...\"],      \u002F\u002F guideline ids actually scored this run (default: all 25)\n  \"scoreboard\": {\n    \"pass\": 4, \"partial\": 8, \"fail\": 11, \"na\": 2,   \u002F\u002F guideline counts by status\n    \"blocked\": 0,                                    \u002F\u002F guidelines gated by a cross-cutting blocker (↺)\n    \"weighted\": { \"points\": 31, \"total\": 142, \"pct\": 22 }  \u002F\u002F summed sub-check weights; pct = points\u002Ftotal\n  },\n  \"blockers\": [                          \u002F\u002F cross-cutting issues gating ≥3 guidelines (may be empty)\n    { \"id\": \"A\", \"title\": \"WAF challenges agent fetchers\", \"gates\": [\"m1-1\", \"m1-3\", \"m2-2\"] }\n  ],\n  \"guidelines\": [\n    {\n      \"id\": \"m1-1\",                      \u002F\u002F matches content\u002Fm1-1-*.md and audit\u002Fm1-1.md\n      \"title\": \"Discovery files\",        \u002F\u002F the audit rubric's short title\n      \"status\": \"partial\",              \u002F\u002F \"pass\" | \"partial\" | \"fail\" | \"na\" | \"blocked\"\n      \"points\": 2,                       \u002F\u002F sub-check weights earned\n      \"weight_total\": 10,                \u002F\u002F sum of the rubric's sub-check weights (matches audit frontmatter)\n      \"complexity\": 1,                   \u002F\u002F 1-5, from the rubric\u002Fcontent frontmatter\n      \"impact\": 4,                       \u002F\u002F 1-5, from the rubric\u002Fcontent frontmatter\n      \"priority\": 20,                    \u002F\u002F impact × (6 - complexity); higher = fix first\n      \"visual_change\": \"none\",          \u002F\u002F \"none\" | \"low\" | \"medium\" | \"high\"\n      \"sub_checks\": [                    \u002F\u002F one entry per rubric row, in order\n        { \"name\": \"sitemap.xml exists\",     \"pass\": true,  \"weight\": 1, \"evidence\": \"200 application\u002Fxml\" },\n        { \"name\": \"Content-Signal present\", \"pass\": false, \"weight\": 1, \"evidence\": \"no Content-Signal: line\" }\n        \u002F\u002F weight: 0 rows are manual\u002Fbonus checks - surfaced but never drag the automated score\n      ],\n      \"fix_summary\": \"Add Content-Signal directive; create llms.txt and index.md; emit Link: headers.\"\n    }\n    \u002F\u002F ... one object per guideline in scope\n  ]\n}\n```\n\n**Field notes.** `weighted.total` counts only guidelines that were scored (N\u002FA and blocked guidelines are excluded), so `pct` reflects what was actually testable. `status` maps from `points \u002F weight_total` per the rubric's thresholds (default **Pass ≥ 85%**, **Partial ≥ 30%**, **Fail \u003C 30%** - some rubrics override this; see [`audit\u002FREADME.md`](.\u002Faudit\u002FREADME.md)). A sub-check with `\"weight\": 0` is a `(manual)` or bonus row: it appears for visibility but can never lower the automated score. `priority` is the same `impact × (6 - complexity)` ranking the action plan uses.\n\n## Cite it\n\nLicensed CC BY 4.0 - use, adapt, and quote freely with attribution to Forter.\n\n## Contribute\n\nSpotted an outdated reference? Missing protocol? Better wording for a guideline? PRs welcome.\n\n1. Edit the relevant markdown file in `content\u002F`.\n2. Keep the frontmatter and H1 in sync (see [`CONTRIBUTING.md`](.\u002FCONTRIBUTING.md) for the schema).\n3. Open a PR against `main`.\n\nThe rendered PDF and webinar are produced from a separate build pipeline. CI (`npm test`) validates frontmatter, audit↔content alignment, and required section structure on every PR - so you'll find out before merge if something is off.\n\n## Structure\n\n```\ncontent\u002F                    Human-facing guide (markdown, the editorial source of truth)\n  00-toc.md                 Table of contents\n  01-introduction.md        Lifecycle frame: Discover → Comprehend → Trust → Act → Experience\n  m{1-5}-0-module-*.md      Five module overviews\n  m{1-5}-{1-N}-*.md         Twenty-five guideline files\n\naudit\u002F                      Machine-testable rubrics - one per guideline, used by SKILL.md\n  m{1-5}-{1-N}.md           Probe + weighted sub-checks + codebase hints\n  README.md                 Rubric file format & strict scoring rules\n\nSKILL.md                    Claude Code skill - orchestrates probes and writes report\u002FAUDIT.md\nCONTRIBUTING.md             Frontmatter schema, body-section checklist, PR template\nscripts\u002Fvalidate.mjs        Self-contained frontmatter + alignment validator (npm test)\nLICENSE                     CC BY 4.0\n```\n\n## License\n\nContent is licensed [CC BY 4.0](.\u002FLICENSE). Built and maintained by [Forter](https:\u002F\u002Fwww.forter.com).\n","该项目提供了一份实用且有观点的指南，旨在帮助任何网站变得对代理友好，即易于发现、理解、信任、操作和体验，从而让基于LLM的代理（及其背后的人类）能够找到、理解、信任并采取行动。核心功能包括五个模块共25条指导原则，每条原则都包含“是什么及为什么”、1-5复杂度\u002F影响评分、具体步骤和参考资料。适用于希望提高其网站对自动化代理可见性和交互性的开发者或团队。此外，项目还提供了一个自动审计工具，可以针对网站进行评估，并根据结果提出改进建议。",2,"2026-06-11 04:10:18","CREATED_QUERY"]