Merge pull request #179 from PostHog/fury-road-remove-llm-prompts

delete llm-prompts directory, move og posthog integration instruction…

Author: edwinyjlim

README.md

@@ -46,8 +46,7 @@ examples/
 │   ├── react-tanstack-router-code-based/   # React with TanStack Router (code-based)
 │   ├── tanstack-start/          # TanStack Start
 │   └── django/                  # Django
-├── llm-prompts/                 # Workflow guides for AI agents
-├── mcp-commands/                # MCP command prompts (`/command` in agents, can wrap `llm-prompts`)
+├── mcp-commands/                # MCP command prompts (`/command` in agents)
 └── scripts/                     # Build scripts
 ```
 
@@ -68,14 +67,14 @@ Run `npm run build:docs` to generate:
 ### Manifest structure
 
 The manifest defines:
-- **Workflows**: Step-by-step guides with automatic next-step linking
+- **Skills**: Bundled skill packages, each containing a `SKILL.md`, `references/` step files, and any docs/example content
 - **Docs**: PostHog documentation URLs (fetched at runtime)
 - **Prompts**: MCP command prompts with template variable substitution
 - **Templates**: Resource templates for parameterized access (e.g., `posthog://examples/{framework}`)
 
 ### Adding new resources
 
-**Workflows**: Add markdown files to `llm-prompts/[category]/` following the naming convention `[order].[step]-[name].md`
+**Skill step files**: Add numbered markdown files to `transformation-config/skills/<skill>/references/` following the convention `<n>-<name>.md` with `next_step:` frontmatter pointing to the next file.
 
 **Examples**: Add new example projects to `basics/` and configure in `scripts/build-examples-mcp-resources.js`
 
@@ -90,8 +89,6 @@ The build script automatically discovers, orders, and generates URIs for all res
 - **Easy to extend**: Add resources by creating properly named files
 - **Version controlled**: Resources evolve with the examples
 
-See `llm-prompts/README.md` for detailed workflow conventions.
-
 ## Security scanning
 
 Before we ship any skills, we run them through [the warlock](https://github.com/PostHog/warlock), PostHog's security scanner for agentic flows. It reads the built skill bundles and looks for prompt-injection attempts and other risky content that could trick an agent downstream. An LLM triage pass then sorts the real threats from the false positives, so we're not chasing noise.

llm-prompts/README.md

@@ -45,7 +45,6 @@ async function main() {
     const distDir = path.join(repoRoot, 'dist');
     const skillsDir = path.join(distDir, 'skills');
     const tempDir = path.join(distDir, 'skills-temp');
-    const promptsDir = path.join(repoRoot, 'llm-prompts');
 
     try {
         fs.mkdirSync(skillsDir, { recursive: true });
@@ -54,7 +53,6 @@ async function main() {
             repoRoot,
             configDir,
             outputDir: tempDir,
-            promptsDir,
             version: BUILD_VERSION,
         });
 

mcp-commands/posthog-setup.json

@@ -49,7 +49,6 @@ const repoRoot = path.join(import.meta.dirname, '..');
 const configDir = path.join(repoRoot, 'transformation-config');
 const distDir = path.join(repoRoot, 'dist');
 const skillsDir = path.join(distDir, 'skills');
-const promptsDir = path.join(repoRoot, 'llm-prompts');
 const skillsSourceDir = path.join(configDir, 'skills');
 const basicsDir = path.join(repoRoot, 'basics');
 
@@ -105,7 +104,6 @@ async function runPartialRebuild(ids) {
         repoRoot,
         configDir,
         distDir,
-        promptsDir,
         version: BUILD_VERSION,
         docContents,
         log: console.log,

scripts/build.js

@@ -241,7 +241,6 @@ async function partialRebuild({
     repoRoot,
     configDir,
     distDir,
-    promptsDir,
     version,
     docContents,
     log = console.log,
@@ -259,7 +258,6 @@ async function partialRebuild({
             repoRoot,
             configDir,
             outputDir: tempDir,
-            promptsDir,
             version,
         });