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/jayminwest/mulch/agents-mdgit clone --depth 1 https://github.com/jayminwest/mulchWhat 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 | $0.02678 | $0.02678 |
| Opus 5 | $0.01339 | $0.01339 |
| Sonnet 5 | $0.00536 | $0.00536 |
| Haiku 4.5 | $0.00268 | $0.00268 |
Grade A, and why
mulch 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 2d 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 — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
This file is the canonical entry point for AI coding agents working in
the mulch repo, following the agents.md convention.
It mirrors the essentials from CLAUDE.md; when the two disagree,
CLAUDE.md is authoritative and this file should be updated to match.
What this project is
Mulch is a passive CLI (@os-eco/mulch-cli) that manages structured
expertise files for coding agents. It has no LLM dependency — agents
call ml record to store an insight and ml prime to load relevant
expertise back into context, and mulch handles storage, validation, and
retrieval. The JSONL files on disk are the database; ml prime renders
them as agent-optimized markdown. Bun is the runtime: the source .ts
files execute directly with no build step.
Mulch is part of the os-eco ecosystem alongside warren (control
plane), burrow (sandbox), plot (coordination), seeds (issue tracking),
and canopy (prompt management). See CLAUDE.md for the long-form design
notes and ROADMAP.md for direction.
Tech stack at a glance
- Runtime: Bun (runs TypeScript directly; no build step on the CLI).
- Language: TypeScript with strict mode — no
any, no@ts-ignore, no@ts-expect-error. - Lint / format: Biome (
biome.json). Errors fail CI; warnings are promoted to errors via--error-on-warnings. - Tests:
bun testdiscovers*.test.tsundertest/and next to scripts; configuration lives inbunfig.toml. - Storage: JSONL files under
.mulch/expertise/(one file per domain, append-only,merge=uniongitattribute for conflict-free concurrent writes). - CLI:
ml/mulch(entry pointsrc/cli.ts, dispatched via commander).
Project layout
mulch/
├── src/
│ ├── cli.ts # ml / mulch CLI entry point + Commander wiring + VERSION
│ ├── index.ts # library entry point
│ ├── api.ts # programmatic API surface
│ ├── commands/ # one file per subcommand (record, prime, sync, ...)
│ ├── schemas/ # record schemas + JSON schema definitions
│ ├── registry/ # type registry (built-in + custom types)
│ └── utils/ # lock, expertise IO, git/worktree, runtime flags
├── scripts/ # quality-gate + report scripts and their budgets
│ ├── check-all.ts # canonical quiet gate runner (byte-identical fleet-wide)
│ ├── check-ci-parity.ts # CI <-> check:all parity detector (byte-identical fleet-wide)
│ ├── ci-parity-config.json # per-repo parity escape hatches (aliases / ciOnly)
│ ├── validate-agents-md.ts # validates this file's references
│ ├── check-file-sizes.ts
│ ├── check-debt-markers.ts
│ ├── check-coverage.ts
│ ├── report-test-timing.ts
│ ├── report-quality-metrics.ts
│ └── version-bump.ts
├── test/ # bun tests mirroring src/ layout
├── .mulch/ # mulch's own expertise (dogfood)
├── .factory/skills/ # repo-local agent skills
├── .github/workflows/ # ci.yml + publish.yml + auto-merge.yml
├── README.md # user-facing pitch
├── CHANGELOG.md # release history
├── RUNBOOK.md # release / triage / rollback procedures
├── CONFIG.md # mulch.config.yaml reference
├── ROADMAP.md # direction
├── biome.json
├── bunfig.toml
├── tsconfig.json
└── package.json
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.
- 2d ago First seen · 225 lines · 2,678 tokens per session scan A db5897db31ea
mulch AGENTS.md is an instructions file published in the GitHub repository jayminwest/mulch (333 stars, last pushed 21d ago), licensed MIT. It adds 2,678 tokens to every session, about $0.0134 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 instructions, from other repositories
wife AGENTS.md
Instructions for ma-nucho-pro/wife, covering if the user wants to install it, if the user wants to change something and rules that are not negotiable.
mnemo-mcp AGENTS.md
Instructions for sshisto/mnemo-mcp, covering agents.md - mnemo-mcp, build / lint / test commands, run a single test file, run a single test class and run a single test function.
SiftMarks CLAUDE.md
Instructions for Lling0000/SiftMarks, covering claude.md, commands, architecture, the four entry points share one database and package layering (lower depends on higher).
task-orchestrator CLAUDE.md
Instructions for jpicklyk/task-orchestrator, covering project: mcp task orchestrator, build commands, docker (most common), architecture and modes of operation.
Project-AMBER AGENTS.md
Instructions for JJHbrams/Project-AMBER, covering agent orchestration, roles and non-negotiable rules.
SiftMarks AGENTS.md
Instructions for Lling0000/SiftMarks, covering siftmarks agent notes, project shape, commands, current caveats and working rules.