Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/seabeardev/reverie/agents-mdgit clone --depth 1 https://github.com/seabearDEV/ReverieWrote 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/instructions/seabeardev/reverie/agents-md)<a href="https://agentmods.dev/instructions/seabeardev/reverie/agents-md"><img src="https://agentmods.dev/badge/instructions/seabeardev/reverie/agents-md.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00862 | $0.00862 |
| Opus 5 | $0.00431 | $0.00431 |
| Sonnet 5 | $0.00172 | $0.00172 |
| Haiku 4.5 | $0.00086 | $0.00086 |
Grade B, and why
Reverie AGENTS.md scanned grade B with 1 finding 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 4d 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.
Reads MCP configurationmediumAgent snooping
mcp.json carries server URLs and auth tokens; reading it lets a mod discover and abuse other integrations.
- **MCP server** (`rvr mcp-server`) — for shell-less surfaces only (e.g. Claude Cowork/Desktop), registered at the app level. Do not add project-level MCP config (`.mcp.json`); shell agents use the CLI. How it starts
The opening of the file, as written. The whole thing — 52 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reverie — agent guide
Reverie store
This project keeps durable, cross-session knowledge in a Reverie store (.reverie/). Fleet policy: transport follows capability.
rvrCLI — the uniform surface for every shell-capable agent (Claude Code, Codex, …):rvr context,rvr get,rvr set,rvr find, … Pipe--jsonoutput (e.g. throughjq) so only the distilled answer enters context. Runrvr manifest --jsonfor the full command map andrvr config llm-instructions --surface clifor CLI specifics (the--jsonenvelope, scope flags, confirm flow,RVR_SESSION).- MCP server (
rvr mcp-server) — for shell-less surfaces only (e.g. Claude Cowork/Desktop), registered at the app level. Do not add project-level MCP config (.mcp.json); shell agents use the CLI.
Attribution: each agent's harness sets RVR_AGENT_NAME (sol = Codex, fable = Claude Code) so audit/telemetry rows identify the writer.
CLI commands emit a single machine-readable envelope with --json (or RVR_OUTPUT=json): { "reverie": "1", "ok": true, "command": "…", "result": …, "error": { "code": … }, "warnings": [] }. Branch on error.code, not on prose. (One exception: rvr audit --follow --json streams raw NDJSON rows, no envelope.)
Never hand-edit .reverie/*.json directly — that bypasses audit logging, alias resolution, interpolation, and staleness metadata (see conventions.editSurface in the store). Use rvr (or, on shell-less surfaces, the MCP tools) only.
Bootstrap
- Load all stored project knowledge first —
rvr context. - If this is a git repo, run
gh issue list --state openand read any issue related to the user's request before coding.
Before exploring code
Check the store before globbing/grepping or reading code — rvr get <key>:
files.<name>before searching for a source file.arch.<area>before reading code to understand a subsystem.conventions.<topic>before making style/pattern decisions.
Write back
When you discover something non-obvious (a gotcha, an architectural decision, a pattern), store it — rvr set <key> <value> — before the session ends. At session end, consider a context.next_session handoff note for the next session's bootstrap banner.
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.
- 4d ago First seen · 52 lines · 862 tokens per session scan B 4ee59c98fe52
Reverie AGENTS.md is an instructions file published in the GitHub repository seabearDEV/Reverie (1 stars, last pushed 10d ago), licensed MIT. It adds 862 tokens to every session, about $0.0043 per session on Opus 5. A static security scan graded it B with 1 finding (reads mcp configuration). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
swag-mcp CLAUDE.md
Claude Code instructions for jmagar/swag-mcp, covering swag mcp - claude memory reference, project overview, core architecture, key components and common development commands.
state-memory-mcp copilot-instructions.md
Instructions for putervision/state-memory-mcp, covering state memory (state-memory-mcp), 1. priority order, 2. when to write to the graph, 3. workflow pattern and 4. codebase seeding on initialization.
toon-memory AGENTS.md
Instructions for LuiggiVal08/toon-memory, covering agents.md, build & verify, architecture, smarter memory pattern and visual memory pattern.
clawstash CLAUDE.md
Claude Code instructions for fo0/clawstash, covering claude.md -- project guide, session start -- read order, workflow triggers, output languages and performance / modes.
immichctrl AGENTS.md
Instructions for lidless-labs/immichctrl, covering agents.md - working in this repo, every session, definition of done, memory owner and memory handoff (mandatory).
immichctrl CLAUDE.md
Instructions for lidless-labs/immichctrl, covering claude.md - claude code rules, project rules, memory handoff, brigade work loop (mandatory) and closeout.