docs: document ThreatLocker extension response shape + add AI topic index

[maximelb]

What

Two discoverability improvements found during routine review of the docs, aimed at automation and AI consumers:

1. ThreatLocker extension — Response shape. The extension page documented each action's request body but not the response envelope a consumer receives. Added a concise "Response shape" subsection to the approval-request reads:

   • The Portal payload is returned under a top-level data key — an array for the search/list read (approval_request_search), a single object for the single-id read (approval_request_get).
   • Documented the salient approval-request record fields an automation depends on (approvalRequestId, statusId with 1=pending, hostname, username, requestorEmailAddress, requestorReason (base64), computerId, organizationId, path, dateTime) plus the nested threatLockerActionDto (actionType, sha256, fullPath, certs, osType, processName).
   • Clarified that an empty data array is a successful empty result, not an error.
   • Used placeholders only (no tenant-specific values), and pointed at the already-cited public ThreatLocker Portal Swagger as the authoritative field-set reference.

2. AI topic index. Added docs/index-for-ai.md: a short, curated map of the top-level sections and — importantly — the adapter-vs-extension dual layout, with one line per integration that has both, pointing to BOTH its adapter path and its extension path (currently ThreatLocker). It carries a header note that it is curated and may lag the tree, and complements the existing auto-generated /llms.txt (a flat nav dump that cannot express these cross-references). Wired into the Reference nav.

Notes

• The brief mentioned a possible dangling internal REFERENCES.md pointer in the ThreatLocker docs. None exists — both pages already point readers to the public Portal Swagger for field-set details, and the new Response shape section reinforces that public pointer.
• Followed the repo's existing llms.txt convention (generated by hooks/llms_txt.py, never committed) by adding the curated companion as index-for-ai.md rather than a static llms.txt that would collide with the generated one.

Validation

• npx markdownlint-cli2 docs/**/*.md — 0 errors.
• mkdocs build — clean (only the generic Material/MkDocs advisory banner); new page builds, no broken relative links introduced, index-for-ai.md present in nav.

🤖 Generated with Claude Code