Borrowing it
Nothing to install: this file belongs to cq27-dev/rag-rat. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/cq27-dev/rag-rat/main/.agents/skills/configure-rag-rat-dream/SKILL.mdgit clone --depth 1 https://github.com/cq27-dev/rag-ratWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cq27-dev/rag-rat/configure-rag-rat-dream)<a href="https://agentmods.dev/skills/cq27-dev/rag-rat/configure-rag-rat-dream"><img src="https://agentmods.dev/badge/skills/cq27-dev/rag-rat/configure-rag-rat-dream/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/cq27-dev/rag-rat/configure-rag-rat-dream"><img src="https://agentmods.dev/badge/skills/cq27-dev/rag-rat/configure-rag-rat-dream.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Rogue Agent · line 73 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
- medium Rogue Agent · line 118 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00177 | $0.01571 |
| Opus 5 | $0.00088 | $0.00785 |
| Sonnet 5 | $0.00035 | $0.00314 |
| Haiku 4.5 | $0.00018 | $0.00157 |
Grade A, and why
configure-rag-rat-dream scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 129 lines — stays where its author put it; the contents beside it link to each section on GitHub.
configure-rag-rat-dream — enable and run the AI memory-maintenance passes
rag-rat's dream command has an optional model turn: the --verify / --compact passes that
AI-verify repo memories (flag those whose claims no longer hold → memory_reality) and compact
them (shorter drive-by summaries). It is rag-rat's only generative-model dependency, off by
default, and gated by a deterministic layer — rag-rat dream stays 100% deterministic unless
[llm.dream] enabled = true and you pass --verify/--compact.
It earns its keep only once a repo has accumulated memories, so this is a later, deliberate
setup, not something to do at first index. The dream-review skill triages the findings these passes
produce — run this skill to turn the model on, that one to work through what it surfaces.
When to set it up
- rag-rat has been in use a while and memories have accumulated.
- The user wants AI verify (reality-check memories) and/or compact (shorter summaries).
Otherwise skip it — the deterministic dream is enough.
Config — mirrors [llm.embedding.remote], but a CHAT backend
Same Connect / Ephemeral split as embeddings, but the model is a small chat LLM served over
/v1/chat/completions. backend must be ollama or vllm — infinity is embed-only and is
rejected here. (If the repo already runs a local infinity for embeddings, that is NOT reusable for
dream — dream needs a generation server.)
Connect (a chat server you already run):
[llm.dream]
enabled = true
[llm.dream.remote]
backend = "ollama"
endpoint = "http://localhost:11434"
model = "qwen3:4b-instruct" # server-side (ollama) name
request_timeout_s = 300
Ephemeral (a Modal/RunPod box — same cookbook as embeddings). CPU inference on a dense evidence pack is pathologically slow (it can blow past the timeout and never verify), so a GPU is the practical option:
[llm.dream]
enabled = true
[llm.dream.remote]
backend = "vllm" # vllm | ollama (NOT infinity)
cookbook = "@rag-rat/cookbook modal" # or @rag-rat/cookbook runpod
gpu = "L4" # 24 GB fits a 4B in fp16 — same class as the embedding box
model = "Qwen/Qwen3-4B-Instruct-2507" # HF id for vllm; the memory-compaction eval winner
auth_env = "MODAL_TOKEN" # optional: env var holding the box's bearer token
request_timeout_s = 900 # a dense evidence pack is slow to first token
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 129 lines · 177 tokens per session scan A 1d18a1c2ad4b
configure-rag-rat-dream is a skill published in the GitHub repository cq27-dev/rag-rat (20 stars, last pushed 4d ago), licensed MIT. It adds 177 tokens to every session and 1,571 once invoked, about $0.0009 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
vault-for-llm
Connect OpenClaw to Vault Agent Memory as a local-first governed project memory layer. Search first, then bounded-read cited source ranges; propose new memories as candidates instead of writing directly into active memory.
skills
Your next session starts cold. No memory of what you built, what broke, what you decided. Every memory you write is a gift to that future session. The richer the memory, the less time re-learning.
memem-mine
Opt in to memem's event-triggered mining. New sessions are mined automatically via the Stop hook.
any-mcp
Use when an agent needs to read, search, create, edit, organize, upload, or chat with Anytype through the bounded any-mcp server. Covers safe tool selection, exact identifier chaining, optional toolsets, common PKM workflows, and the narrow anyr fallbacks for chat subscriptions and rich chat blocks.
stats
Show Captain Memo's corpus statistics (chunks per channel, observation counts, indexing progress, embedder info). Use when the user types /captain-memo:stats.
observations
List recent captured session observations (the Haiku-summarized voyage logs). Use when the user wants to see what Captain Memo has logged from past sessions.