AI Features

AI assistants, agent adapters, skills, public AI artifacts, and automated AI-facing pipelines in the docs repo.

AI Features

The AI layer has two audiences: readers who consume AI-ready docs artifacts, and contributors or agents who use repo-native adapters, skills, and governance workflows. The ownerless standard applies to both: AI outputs must be generated from canonical sources, and AI agents must follow the same gates as humans.

Public AI Surfaces

Surface	Purpose	Canonical source	Repair or generation path
llms.txt	AI-first root artifact for crawlers and assistant ingestion.	docs.json, docs content, operations/scripts/generators/ai/llm/generate-llms-files.js	node operations/scripts/generators/ai/llm/generate-llms-files.js --write
sitemap-ai.xml	AI-oriented sitemap for docs retrieval.	docs.json, operations/scripts/generators/content/seo/generate-ai-sitemap.js	lpd ai-sitemap --write
Mintlify AI assistant	Reader-facing assistant backed by published docs.	Published Mintlify docs site	Fix source pages and redeploy, not assistant answers directly.
Mintlify MCP	External tool access to published docs.	Mintlify project configuration	Validate published docs and route coverage.

Agent Adapters

Agent	Repo adapter	Role
Codex	.github/AGENTS.md, .codex/task-contract.yaml	Task isolation, branch contracts, staged scope validation, no-stash safety.
Claude Code	.claude/CLAUDE.md, .claude/skills/	Project context, canonical skills, session workflows.
Cursor	.cursor/rules/repo-governance.mdc, .cursor/rules/no-deletions.mdc	Governance and deletion guardrails.
Windsurf	.windsurf/rules/repo-governance.md	Governance and repo safety rules.
Augment	.augment/rules/	Governance, git safety, and deletion guardrails.
GitHub Copilot	.github/copilot-instructions.md	Inline coding guidance.

All adapters inherit the root baseline in AGENTS.md. Adapter-specific files should only add tool-specific behavior, not contradict root governance.

Skills and Agent Workflows

There are 34 local skills under ai-tools/ai-skills/. They fall into seven practical groups:

Group	Skills	Purpose
Session and process	thread, pm, research, design, build, iterate, close, dispatch, agent-brief, diagnose, propagate	Plan, research, design, execute, review, and close work with evidence.
Content and docs	docs-copy, page-authoring, content-pipeline-pass-a, content-pipeline-pass-b, content-pipeline-tab-map	Produce and review docs using the content pipeline.
Review and audit	repo-audit-orchestrator, docs-coverage-and-route-integrity-audit, docs-quality-and-freshness-audit, rubric-static-review, script-footprint-and-usage-audit	Find structural, content, and governance issues.
Governance creation	create-action, create-component, create-script, skill-docs, generated-mdx-banners-governance	Scaffold governed artifacts with required metadata.
Component governance	component-layout-governance, create-component	Keep UI primitives aligned with the component framework.
Cleanup	cleanup-quarantine-manager, script-footprint-and-usage-audit	Classify, quarantine, and retire stale artifacts through approved paths.
Agent packaging	cross-agent-packager, browser, product-thinking, style-and-language-homogenizer-en-gb	Package skills and support specialized execution.

AI Governance Contracts

Contract	Why it matters	Validation surface
Agent baseline freshness	All AI adapters should reflect the same core repo rules.	operations/scripts/validators/governance/compliance/check-agent-docs-freshness.js
Codex task isolation	AI task branches must declare scope and avoid overlapping or hidden work.	.codex/task-contract.yaml, operations/scripts/validators/ai/codex/validate-locks.js
Skill registry parity	Skills and generated agent packs should remain synchronized.	ai-tools/registry/, operations/scripts/generators/governance/catalogs/generate-ai-skills-indexes.js
Public AI artifacts	AI-facing root files must be generated and validated from current nav/content.	lpd ai-sitemap --check, llms file validators

AI Pipeline Risks

Risk	Current evidence	Required fix
Skill count drift	docs-guide/features/ai-features.mdx previously documented 28 skills; live inventory has 34.	Use generated skill indexes or cite live inventory rather than hand-maintained counts.
Old validator references	Some guides still point at missing or renamed validator paths.	Route AI-tool checks through current operations/scripts/generators/governance/catalogs/* and compliance validators.
Agent adapter duplication	Multiple adapters repeat governance rules.	Keep root policy in AGENTS.md and make adapters thin, tool-specific extensions.
AI artifact drift	llms.txt, sitemap-ai.xml, and docs indexes can become stale after nav or content moves.	Regenerate in the same change as route/content movement and run staged validation.

Contributor Rules

• Do not edit generated AI artifacts by hand.
• Do not add a new skill without a SKILL.md, registry path, and validation story.
• Do not treat AI-generated reports as source of truth unless their generator and source inputs are named.
• Do not bypass branch or deletion rules for AI work.

Related

• docs-guide/tooling/ai-tools.mdx
• docs-guide/policies/agent-governance-framework.mdx
• docs-guide/policies/ownerless-governance.mdx
• ai-tools/registry/ai-tools-inventory.md