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 skills add NexaDuo/mARC --skill initgit clone --depth 1 https://github.com/NexaDuo/mARCWrote 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/nexaduo/marc/init)<a href="https://agentmods.dev/skills/nexaduo/marc/init"><img src="https://agentmods.dev/badge/skills/nexaduo/marc/init/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/nexaduo/marc/init"><img src="https://agentmods.dev/badge/skills/nexaduo/marc/init.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00122 | $0.03354 |
| Opus 5.5 | $0.00049 | $0.01342 |
| Sonnet 5.5 | $0.00024 | $0.00671 |
| Haiku 4.5 | $0.00012 | $0.00335 |
Grade A, and why
init 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 1mo 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 — 290 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/marc:init — opt-in onboarding & config scaffolding
You are running the mARC onboarding flow. Your job is to help the user turn a
zero-config repo into one with a persistent, versioned team binding, so
@techlead and the specialists stop relying on ephemeral session memory and read
the repo's concrete facts from {{ agents_dir }}/team.toml (and optionally AGENTS.md)
at the start of every session.
The one rule that overrides everything
Nothing is ever written silently. For every artifact you produce you MUST:
- Discover / compose the exact content.
- Show that exact content to the user (fenced, verbatim).
- Ask for an explicit "yes" for that specific artifact.
- Write only on that confirmation, to the path shown.
There are three independent, individually opt-in artifacts. The user may accept any subset (including none). Never bundle them into one confirmation. Never infer approval from silence or from a "yes" to a different artifact. If the user declines everything, the repo is left byte-for-byte unchanged and zero-config behavior is fully preserved.
Do not invent facts. Anything you cannot discover empirically becomes a
clearly-marked TODO placeholder, never a guess.
Step 0 — Discover the repo facts at runtime (no hardcoded values)
Mirror the tech-lead skill's runtime-discovery pattern. Resolve values from gh
against the checked-out repo; do not hardcode any org, repo slug, or project
number in this flow.
Batch the probes into ONE block. Run all discovery in a single Bash
invocation so onboarding doesn't fire a permission prompt per gh call (a dogfood
run fired ~7). This block is read-only discovery — the Write confirmations
below (steps 2-4 of the one rule) are the intentional safety gate and are NOT
what we are reducing; every file write still stops for an explicit "yes".
# Where onboarding writes (the CONSUMING repo, not the plugin):
ROOT="${{{ project_dir_env }}:-$PWD}"
# --- ORG + REPO (from the checked-out repo) ---
GH_REPO="$(gh repo view --json nameWithOwner -q .nameWithOwner 2>/dev/null)"
GH_ORG="${GH_REPO%%/*}"
# --- PROJECT (v2) candidates for the org, if any ---
# Needs the `project` scope; if it errors with "missing required scopes", tell the
# user to run `gh auth refresh -s project,read:project` once, then retry. Do not
# fabricate a number — leave it as a TODO if it can't be read. LIST candidates
# (number + title); the guard below decides — never auto-pick .projects[0].
echo "== org/repo: ${GH_REPO:-<unresolved>} =="
gh project list --owner "$GH_ORG" --format json 2>/dev/null \
| jq -r '.projects[] | "\(.number)\t\(.title)"'
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.
- 1mo ago First seen · 290 lines · 122 tokens per session scan A d6ab7d33c5ca
init is a skill published in the GitHub repository NexaDuo/mARC (6 stars, last pushed today), licensed MIT. It adds 122 tokens to every session and 3,354 once invoked, about $0.0005 per session on Opus 5.5. 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 skills, from other repositories
bug-sweep
Codebase bug hunt: fix AI-fixable bugs, defer the rest.
review
Review a plan, diff, or roadmap slice; findings land on it.
dogfood
Fix-through loop: invoke a new thing, fix until it works.
enrich-and-review
Runs the enrichment pipeline over plan chunk directories.
systematic-debugging
Root-cause one known bug: reproduce, trace, fix, verify.
brainstorming
Shape vague or multi-subsystem asks into requirements. Size first if unsized.