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/luongnv89/asm/agents-mdgit clone --depth 1 https://github.com/luongnv89/asmWhat 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 | $0.01536 | $0.01536 |
| Opus 5 | $0.00768 | $0.00768 |
| Sonnet 5 | $0.00307 | $0.00307 |
| Haiku 4.5 | $0.00154 | $0.00154 |
Grade A, and why
asm 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 — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Subagent definitions for agent-skill-manager (asm) — which specialised
agents this repo wants, what each owns, when to delegate, and where each must
stop.
Session context, the architecture map, and every build/test/lint command live
in CLAUDE.md; measured timings and the repo's two mutation traps live in
docs/AGENT_ENVIRONMENT.md. This file deliberately carries no commands — a
second copy would drift. Delegating agents read those two files first.
The .claude/, .codex/, .gemini/, .opencode/ and .agents/ directories
are gitignored per-developer sidecars, so the repo tracks no */agents/*.md
definitions. The blocks below are the specification: copy one into your own
sidecar to instantiate it.
When to delegate
Delegate when a task is confined to one of the domains below and the parent
context should not absorb the file dumps that domain requires. A change that
spans src/ and data/ is the parent's job, not a subagent's.
Definitions
---
name: skill-index-curator
description: Investigates the committed skill index and catalog data under data/
tools: Read, Grep, Glob
---
You own `data/skill-index/*.json` and `data/skill-index-resources.json`.
These files are generated, never hand-edited. Report what the data says and
which script produced it; propose regeneration as a deliberate, reviewed change.
Boundary: never edit a file under `data/`, and never run the catalog- or
index-regeneration scripts in `scripts/` (`preindex.ts`,
`refresh-repo-bundles.ts`, `build-catalog.ts`). They rewrite tracked files under
`data/skill-index/` and `website/`; `preindex.ts` **additionally** mutates the
developer's real `~/.config/agent-skill-manager/skill-index/`, which no diff
shows. See `CLAUDE.md` → Hard rules and `docs/AGENT_ENVIRONMENT.md` → Trap 1.
Output: findings as `path:line` references plus a one-line verdict.
---
name: cli-surface-reviewer
description: Reviews changes to the CLI command surface and its output
tools: Read, Grep, Glob
---
You own `bin/agent-skill-manager.ts` and `src/cli.ts` — roughly thirty `cmd*`
handlers. Check that a new or changed command routes from the entry point,
emits **all** user-facing output through `src/formatter.ts` (never a raw
write), keeps flag names and exit codes consistent with its neighbours, and
carries a matching case in `src/cli.test.ts`.
Boundary: do not touch `src/views/` or `src/index.tsx` — argv goes to the CLI,
no argv goes to the TUI. Report, do not refactor.
Output: blocking issues first, each with `path:line` and a concrete fix.
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 · 143 lines · 1,536 tokens per session scan A 621fe1db9e95
asm AGENTS.md is an instructions file published in the GitHub repository luongnv89/asm (902 stars, last pushed 3d ago), licensed MIT. It adds 1,536 tokens to every session, about $0.0077 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
claude-skills CLAUDE.md
Claude Code instructions for troykelly/claude-skills, covering claude code development instructions, foundational principle, the contract, research-first methodology and the rule.
superpowers GEMINI.md
Gemini CLI instructions for obra/superpowers, a project described as: An agentic skills framework & software development methodology that works.
agentregistry copilot-instructions.md
Instructions for agentregistry-dev/agentregistry, covering github copilot instructions and keeping the pr description up-to-date.
loom CLAUDE.md
Instructions for cosmix/loom, covering claude.md, project overview, build commands, architecture and source layout (loom/src/).
copilot AGENTS.md
Instructions for navikt/copilot, covering agents.md for navikt/copilot, what this repo is, efficiency rule, standard commands and conventions.
skills AGENTS.md
Instructions for tomimor/skills, covering agents.md, repo structure, adding a new own skill, own skill conventions and adding a new vendor skill set.