Interactive RLM cache initialization. Use when: setting up a new project's semantic cache for the first time, or adding a new cache profile. Walks the user through folder selection, extension config, manifest creation, and first distillation pass.
From rlm-factorynpx claudepluginhub richfrem/agent-plugins-skills --plugin rlm-factoryThis skill is limited to using the following tools:
acceptance-criteria.mdassets/rlm_profiles.jsonassets/rlm_summary_cache_manifest.jsonassets/rlm_tools_manifest.jsonevals/evals.jsonevals/results.tsvreferences/acceptance-criteria.mdrequirements.txtresources/distiller_manifest.jsonresources/manifest-index.jsonresources/prompts/rlm/rlm_summarize_general.mdresources/prompts/rlm/rlm_summarize_tool.mdresources/rlm_manifest.jsonscripts/cleanup_cache.pyscripts/distiller.pyscripts/inject_summary.pyscripts/inventory.pyscripts/query_cache.pyscripts/rlm_config.pyGuides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.
Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.
Deploys Linkerd service mesh on Kubernetes with patterns for installation, proxy injection, mTLS, service profiles (retries/metrics), traffic splits (canary), and authorization policies.
This skill requires Python 3.8+ and standard library only. No external packages needed.
To install this skill's dependencies:
pip-compile ./requirements.in
pip install -r ./requirements.txt
See ./requirements.txt for the dependency lockfile (currently empty — standard library only).
Initialize a new RLM semantic cache for any project. This is the first-run workflow — run it once per cache, then use rlm-distill-agent for ongoing updates.
Real-world examples of each config file are in references/examples/:
| File | Purpose |
|---|---|
rlm_profiles.json | Profile registry -- defines named caches and their manifest/cache paths |
rlm_summary_cache_manifest.json | Project docs manifest -- what folders/globs to include and exclude |
rlm_tools_manifest.json | Tools manifest -- scoped to scripts and plugins only |
Before creating anything, gather requirements:
docs/, src/, plugins/).md, .py, .ts).agent/learning/ or config/rlm/plugins, project, tools)rlm_profiles.jsonEach cache is defined as a profile in rlm_profiles.json. This file is located at RLM_PROFILES_PATH or defaults to .agent/learning/rlm_profiles.json. If it doesn't exist, create it:
mkdir -p <profiles_dir>
Create or append to <profiles_dir>/rlm_profiles.json:
{
"version": 1,
"default_profile": "<NAME>",
"profiles": {
"<NAME>": {
"description": "<What this cache contains>",
"manifest": "<profiles_dir>/<name>_manifest.json",
"cache": "<profiles_dir>/rlm_<name>_cache.json",
"extensions": [
".md",
".py",
".ts"
]
}
}
}
| Key | Purpose |
|---|---|
description | Human-readable explanation of the profile's purpose |
manifest | Path to the manifest JSON (what folders/files to index) |
cache | Path to the cache JSON (where summaries are stored) |
extensions | List of string file extensions to include |
The manifest defines which folders, files, and globs to index. Extensions come from the profile config.
Create <manifest_path>:
{
"description": "<What this cache contains>",
"include": [
"<folder_or_glob_1>",
"<folder_or_glob_2>"
],
"exclude": [
".git/",
"node_modules/",
".venv/",
"__pycache__/"
],
"recursive": true
}
echo "{}" > <cache_path>
Scan the manifest against the cache to find uncached files:
python3 ./scripts/inventory.py --profile <NAME>
Report: "N files in manifest, M already cached, K remaining."
For each uncached file:
{
"<relative_path>": {
"hash": "agent_distilled_<YYYY_MM_DD>",
"summary": "<your summary>",
"summarized_at": "<ISO timestamp>"
}
}
"✅ Cached: <path>"Run audit again:
python3 ./scripts/inventory.py --profile <NAME>
Target: 100% coverage. If gaps remain, repeat Step 6 for missing files.
Every summary should answer: "Why does this file exist and what does it do?"
| ❌ Bad | ✅ Good |
|---|---|
| "This is a README file" | "Plugin providing 5 composable agent loop patterns for learning, red team review, triple-loop delegation, and parallel swarm execution" |
| "Contains a SKILL definition" | "Orchestrator skill that routes tasks to the correct loop pattern using a 4-question decision tree, manages shared closure sequence" |
rlm-distill-agent for ongoing cache updatesrlm-curator for querying, auditing, and cleanup.gitignored if they contain project-specific summaries