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/gf-labs/claude-toolbox/cli-capability-auditnpx skills add gf-labs/claude-toolbox --skill cli-capability-auditgit clone --depth 1 https://github.com/gf-labs/claude-toolboxWrote 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/gf-labs/claude-toolbox/cli-capability-audit)<a href="https://agentmods.dev/skills/gf-labs/claude-toolbox/cli-capability-audit"><img src="https://agentmods.dev/badge/skills/gf-labs/claude-toolbox/cli-capability-audit.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 | $0.00104 | $0.02130 |
| Opus 5 | $0.00052 | $0.01065 |
| Sonnet 5 | $0.00021 | $0.00426 |
| Haiku 4.5 | $0.00010 | $0.00213 |
Grade A, and why
cli-capability-audit 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 5d 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 — 188 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLI capability audit
Derive what a CLI actually implements from its dispatcher source, diff that against what its help text and docs claim, fix the gaps, and publish a capability map that records reality.
Use when docs have drifted, when a repo needs an accurate command inventory, or when someone needs to describe a codebase accurately (a README, a summary, a resume bullet) and the existing docs can't be trusted as the source.
The core idea
Three sets per tool, compared in both directions:
| Set | Source | How |
|---|---|---|
| Implemented | Dispatcher source | Static parse — case arms, if-guards, dispatch chains |
| Advertised | <tool> --help |
Run help only, never a subcommand |
| Documented | README, per-tool docs, agent context files | Text scan |
advertised − implemented→ phantom: user follows docs, hits an errorimplemented − advertised→ undocumented: real capability nobody can find
Both directions matter. Checking only "are my commands documented?" misses phantom commands entirely, and phantoms are the higher-severity class.
Never execute subcommands to introspect. In a real toolkit that opens $EDITOR,
launches fzf, or spends money on an API. Drift detection needs no execution — it is a set
operation over two static reads plus --help.
Workflow
Announce: "Using cli-capability-audit to ."
Resolve the script paths first. The audit runs with the working directory set to the
repo being audited, while these scripts ship inside the plugin — a bare
scripts/extract-surface.py will not resolve. Set once, then use $AUDIT throughout:
AUDIT="${CLAUDE_PLUGIN_ROOT:-$CLAUDE_TOOLBOX_ROOT}/skills/cli-capability-audit/scripts"
CLAUDE_PLUGIN_ROOT is the active plugin install location. If it is unset, fall back to
${CLAUDE_TOOLBOX_ROOT}/skills/cli-capability-audit — the skill ships from claude-toolbox.
If neither is set, invoke the scripts by absolute path under the plugin install location.
What ships with it
4 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.
- 5d ago First seen · 188 lines · 104 tokens per session scan A 17ec301abf1c
cli-capability-audit is a skill published in the GitHub repository gf-labs/claude-toolbox (2 stars, last pushed today), licensed MIT. It adds 104 tokens to every session and 2,130 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 2026-08-31.
Other skills, from other repositories
s-continue
Cheaper and faster than /compact. Restores previous session context from Claude Code AND Codex transcripts by reading them directly — no LLM calls, no token cost. Also auto-loads a handoff written by /s-compact, if one exists. Triggers on "s-continue", "restore context", "what was I doing", "pick up where I left off"…
s-compact
Session handoff for the next session — write-side pair of /s-continue. Triggers on "s-compact", "session handoff", "handoff prompt", "hand off", "wrap up the session", "prep for next session". Distills what /s-continue cannot recover: subagent findings, tool-output numbers, and decisions that never entered the…
usage-view
Know exactly what you spent (Claude Code) or how much of your Codex limit you have used (Codex). Interactive HTML dashboard with token usage and, on Claude Code, cost breakdown and 5-hour window timeline across all sessions.
report-limit
Max plan hit the wall? Report your 5h window data — we're mapping the rate limit formula Anthropic won't publish.
setup-git-lite
Disable Claude Code built-in git instructions and inject a curated 280-tok minimum via SessionStart hook. Saves 2,200 tok/session + 1,700 tok/call.
setup-statusline
Live token counter in your CLI. Shows real-time input/output/cache token counts in the Claude Code status bar.