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/serbyte-development/shellby-mcp/audit-logsnpx skills add serbyte-development/shellby-mcp --skill audit-logsgit clone --depth 1 https://github.com/serbyte-development/shellby-mcpWhat 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 | $0.00024 | $0.00355 |
| Opus 5 | $0.00012 | $0.00178 |
| Sonnet 5 | $0.00005 | $0.00071 |
| Haiku 4.5 | $0.00002 | $0.00036 |
Grade A, and why
audit-logs 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.
What it actually says
Agent Command Log
Use this skill to analyze agent-commands.yaml.
Do not load whole log into context. Parse it with Python, Node, Ruby, shell tools, whatever fits the question.
Source of truth for format:
src/server/audit-log.ts
Header
--- # [!|~] TOOL - DURATIONms - N in [/ N out] [- structured] [- max_output_tokens=N] [- HTTP ...] - HH:MM:SS
!= tool/HTTP/connection failure~= call took at least 5 secondsin= tokens from full serialized arguments before log truncationout= model-facing output tokens when captured- final time = local call start time
Tool Bodies
shell_run: shell/request ID, optional cwd, commandshell_poll: shell/request ID, cursorapply_patch: cwd + patch size; patch body retained only on failure- other tools: serialized
args
Limits
- shell command: 2,000 chars
- generic args: 600 chars
- failed patch: 32,000 chars
- failure message: 1,000 chars
Caveats
- Missing
outdoes not mean zero output. - Shell nonzero exit may not produce
!. - Successful tool output bodies are not stored.
- Entries are written when calls complete, so file order is not guaranteed invocation order.
- Timestamps have no date.
- One file may contain multiple sessions.
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 · 50 lines · 24 tokens per session scan A 7b26f4a91c0b
audit-logs is a skill published in the GitHub repository serbyte-development/shellby-mcp (10 stars, last pushed 5d ago), licensed MIT. It adds 24 tokens to every session and 355 once invoked, about $0.0001 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 skills, from other repositories
codexless-release-supervisor
Prepare, validate, and publish Codexless preview/hotfix releases from the canonical household source, including acceptance anti-omission gates, candidate provenance/parity, manifest/build identity, native Windows/macOS fresh-install acceptance from clean public source, safe Git integration, GitHub prerelease…
codexless-browser-repair
Diagnose and temporarily repair Codexless Browser compatibility after a Codex, Chrome Skill, or Browser runtime update when Codexless Browser stopped working. Use for current-version Codexless Browser compatibility drift, not ordinary website bugs, general Browser operation, Codexless install/update work, or long-term…
codex-with-chatgpt
Use ChatGPT (web) as the planning and review brain for Codex coding sessions, while Codex keeps full execution ownership. Use when the user says "使用 Codex with ChatGPT ..." / "Set up Codex with ChatGPT" / "用 ChatGPT 规划", when they ask to connect ChatGPT to the current workspace, disconnect it, or run a task through…
add-widget
Add a new widget (React UI + MCP tool) to this template. Use when asked to "add a widget", "add a new tool with UI", "create a new widget", "scaffold a widget", or "build a new MCP App feature".
create-mcp-tool
Add a plain (text-only) or UI-enhanced MCP tool to this template's server. Use when asked to "add a tool", "add an MCP tool", "create a new tool", "add a server-side tool without a widget", "add a tool with a UI", "add a tool with a widget", or "create a tool and a widget".
threebrowser-studio-mcp
Author, inspect, render, validate, animate, and save Three.js WebGPU projects in the native ThreeBrowser Studio through its real threestudio MCP tools. Use whenever Codex is asked to build or edit a ThreeBrowser Studio scene, material, Blender-style shader or texture graph, animation, lighting setup, camera, or…