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/agenxy/dibs/agents-mdgit clone --depth 1 https://github.com/Agenxy/dibsWrote 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/agenxy/dibs/agents-md)<a href="https://agentmods.dev/instructions/agenxy/dibs/agents-md"><img src="https://agentmods.dev/badge/instructions/agenxy/dibs/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 | $0.03359 | $0.03359 |
| Opus 5 | $0.01680 | $0.01680 |
| Sonnet 5 | $0.00672 | $0.00672 |
| Haiku 4.5 | $0.00336 | $0.00336 |
Grade A, and why
dibs AGENTS.md 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 3d 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 — 231 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md: orientation for agents working on Dibs
Read PHILOSOPHY.md first; it is the decision procedure. This file is the map.
docs/ARCHITECTURE.md is the territory: how it fits together, and the four bug
classes that keep recurring here. If you are an agent using Dibs rather than
changing it, you want SKILLS.md (also served over MCP as dibs://skills).
What this repo is
A Go daemon (dibd) + CLI (dibs) implementing a local coordination service for
fleets of AI agents. Agents connect over MCP. See SPEC.md (living, not frozen,
change it when reality disagrees, and record why).
Layout
| Path | Role |
|---|---|
internal/core/ |
Pure deterministic state machine. No I/O, no goroutines, no wall clock. Apply(op, now) → (result, events). The invariant state == fold(ledger) lives or dies here. |
internal/ledger/ |
Append-only, hash-chained, encrypted JSONL. Truth. |
internal/engine/ |
Single-writer event loop over core. All impure inputs (liveness probes, clocks) are recorded into ops so replay reproduces decisions. |
internal/mcp/ |
MCP surface: tools, resources, subscriptions/listen. Dual-version (2026-07-28 + legacy). |
internal/blobstore/ |
Content-addressed attachment bytes, encrypted at rest. Outside the replay model by design. |
internal/web/ |
Human board (SSE + htmx; one ~320-line script in internal/assets/board.js, no framework or build step). |
cmd/dibd/ |
Daemon + auth gate. cmd/dibs/ |
Rules you must not break
internal/corestays pure. No I/O, notime.Now(), no randomness, no network. If you need an impure input, record it in theOpso replay applies the same decision.- An op is ledgered iff it changed replayable state. The engine ledgers exactly when the serial advanced. Never let those disagree.
- Non-deterministic things live outside the core as derived, rebuildable views. Losing a derived view must not lose coordination state.
- Advisory, not coercive. Declaring work never fails. Don't add blocking semantics.
- The board may WAKE an agent, and may not steer one. Reaching an idle agent so it
can read its own mail is the product: a message service whose recipient must already
be running is a polling API. The one way it does that is
[wake.exec], argv from the operator's config, no shell, nothing an agent said, rate limited, logged. Everything past "you have mail" is still forbidden: no prompt injection, no session management, no deciding what an agent does next. SeeWAKE-MECHANISMS.md, which argued the other way for months and now records why that was wrong. - Honesty in errors. Every error carries a
hintthat tells a drifted agent the corrective call.
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.
- 3d ago First seen · 231 lines · 3,359 tokens per session scan A 754332256c86
dibs AGENTS.md is an instructions file published in the GitHub repository Agenxy/dibs (2 stars, last pushed today), licensed Apache-2.0. It adds 3,359 tokens to every session, about $0.0168 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 2026-08-31.
Other instructions, from other repositories
shift-mcp-server AGENTS.md
AGENTS.md instructions for cyanheads/shift-mcp-server, covering developer protocol, what's next?, core rules, patterns and tool.
shift-mcp-server CLAUDE.md
Claude Code instructions for cyanheads/shift-mcp-server, covering developer protocol, what's next?, core rules, patterns and tool.
sema AGENTS.md
AGENTS.md instructions for emergent-wisdom/sema, covering repository agent guide, development workflow, vocabulary changes, vocabulary review and the improvement loop, per pattern.
tooltrust-scanner CLAUDE.md
Claude Code instructions for AgentSafe-AI/tooltrust-scanner: Use the /browse skill from gstack for all web browsing. Never use mcpclaude-in-chrome tools.
mcp-anything CLAUDE.md
Instructions for gaarutyunov/mcp-anything, covering claude.md — ai agent instructions for mcp-anything, project, module, commands and build.
sema CLAUDE.md
Claude Code instructions for emergent-wisdom/sema: Read and follow AGENTS.md. It is the canonical repository-level workflow for development, vocabulary changes, deployment boundaries, and verification.