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 skills/webcoyote/sandvault/svnpx skills add webcoyote/sandvault --skill svgit clone --depth 1 https://github.com/webcoyote/sandvaultWrote 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/webcoyote/sandvault/sv)<a href="https://agentmods.dev/skills/webcoyote/sandvault/sv"><img src="https://agentmods.dev/badge/skills/webcoyote/sandvault/sv.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 5.1 | $0.00093 | $0.00893 |
| Opus 5 | $0.00046 | $0.00447 |
| Sonnet 5 | $0.00019 | $0.00179 |
| Haiku 4.5 | $0.00009 | $0.00089 |
Grade A, and why
sv scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sandvault Handoff
Hand off the current task to a sandboxed Claude running inside sandvault.
When to use
The user invokes /sv when they want Claude to continue the current work inside a sandvault sandbox.
Steps
1. Summarize the task
Write a clear, actionable summary of what the user wants done. Include:
- What the goal is
- What approach to take (if already discussed)
- What files are involved
- Any decisions already made in this conversation
- The current branch name and any relevant context
2. Write the handoff file
Write the task briefing to /Users/Shared/sv-$USER/tmp/handoff-<repo>.md,
where $USER is the host user and <repo> is the source repo's
basename (e.g. /Users/Shared/sv-jesse/handoff-sandvault.md). This is
the sandvault shared workspace — readable from inside the sandbox via
the mounted /Users/Shared tree, so the sandboxed Claude can read it
at startup. The briefing is not copied into the clone; it lives in
the shared workspace and the sandboxed Claude is pointed at it by path
(see step 4). Nothing is written into the source repo itself.
The handoff file should include:
- A "# Task Handoff" heading
- Task context, approach, decisions, branch, relevant files
- A "## Setup" section: note that gitignored build artifacts (
.venv/,node_modules/, build dirs) won't be in the clone. Check forrequirements.txt,pyproject.toml,package.jsonetc. and instruct the sandboxed Claude to set up the environment first. - A "## What to do" section with a direct instruction
3. Ask the user for confirmation
Show the user:
- The repo that will be cloned
- A brief summary of the task being handed off
4. Launch in a new terminal window
Use the bundled helper to launch sv-clone in a new window of whatever
terminal the user is running (Terminal.app, iTerm2, Ghostty, WezTerm,
kitty, Alacritty, cmux, Warp, with Terminal.app as a fallback):
skills/sandvault/sv/scripts/launch-in-terminal.sh 'sv-clone <repo-path> -- claude -- "Read /Users/Shared/sv-<host-user>/tmp/handoff-<repo>.md and continue the task described there."'
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 6d ago First seen · 75 lines · 93 tokens per session scan A d9fb66b6e20a
sv is a skill published in the GitHub repository webcoyote/sandvault (401 stars, last pushed 3d ago), licensed Apache-2.0. It adds 93 tokens to every session and 893 once invoked, about $0.0005 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 skills, from other repositories
release
Run the local Agent Safehouse release flow: inspect commits since the last published release, propose the next SemVer version and changelog, present a dry-run for confirmation, then update changelog, publish the GitHub release, and publish the stable Homebrew tap when confirmed.
manage-codex-skills
Inventory, classify, route, catalog, relocate, or validate Codex skills and generate a skills menu. Use for skill-library organization or when Codex must find task-relevant skills. Classification, relocation, creation, revision, and promotion require user approval.
bootstrap-codex-workspace
Inspect a Windows or macOS Codex environment and produce a read-only, approval-bound plan for the global AGENTS.md path, user skills root, and root for future projects. Use for workspace setup, bootstrap, or location planning. Do not write outside the repository until the user approves the exact plan ID.
reflect-codex-workflows
Compare evidence from a current or completed project with installed skills and decide whether to reuse, revise, create, or not skill the lesson. Use for workflow reflection or proposed skill upgrades. Do not edit skills without explicit authorization.
maintain-project-ai-records
Initialize and maintain lightweight AI work records in a writable project. Use on the first substantive writable task and after meaningful milestones. Do not create records for read-only inspection or trivial answers.
organize-codex-projects
Choose a location and internal structure for a future Codex project, including focused projects and umbrella projects with durable subprojects. Use when creating or planning a new project. Do not move, rename, or normalize existing projects.