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 KevinZai/commander --skill ccc-doctorgit clone --depth 1 https://github.com/KevinZai/commanderWrote 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/kevinzai/commander/ccc-doctor)<a href="https://agentmods.dev/skills/kevinzai/commander/ccc-doctor"><img src="https://agentmods.dev/badge/skills/kevinzai/commander/ccc-doctor/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/kevinzai/commander/ccc-doctor"><img src="https://agentmods.dev/badge/skills/kevinzai/commander/ccc-doctor.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00062 | $0.06080 |
| Opus 5.5 | $0.00025 | $0.02432 |
| Sonnet 5.5 | $0.00012 | $0.01216 |
| Haiku 4.5 | $0.00006 | $0.00608 |
Grade D, and why
ccc-doctor scanned grade D with 3 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
else REMOTE_VERSION=$(curl -fsSL --max-time 8 "https://raw.githubusercontent.com/KevinZai/commander/main/commander/cowork-plugin/.claude-plugin/plugin.json" 2>/dev/null | node -e 'let s="";process.stdin.on("data",d=>s+=d Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
SETTINGS="$HOME/.claude/settings.json" Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
else REMOTE_VERSION=$(curl -fsSL --max-time 8 "https://raw.githubusercontent.com/KevinZai/commander/main/commander/cowork-plugin/.claude-plugin/plugin.json" 2>/dev/null | node -e 'let s="";process.stdin.on("data",d=>s+=d How it starts
The opening of the file, as written. The whole thing — 339 lines — stays where its author put it; the contents beside it link to each section on GitHub.
$ccc-doctor — Diagnose the Commander install
CC Commander · $ccc-doctor · one probe → paste-ready diagnostic report → the exact next command
When to use / when not to
- Use when Commander skills or hooks misbehave, after an update, or to produce a report for a GitHub issue or support thread.
- Doctor is read-only. Its single write is the consent-gated
autoUpdatetoggle in Step 4, always backed up and verified. - To update the plugin, use
$ccc-update. To clean junk or add settings keys, use$ccc-tuneup. For this repository's vendor submodules, use$ccc-upgrade. $ARGUMENTS:quick(default) runs Steps 1–5;fulladds the source-tree drift checks in Step 6.- In Codex, ask each AskUserQuestion picker as a plain-text question with the same options; the probe runs unchanged and reports Codex rows.
What happens
1. Probe the install and platform (one Bash call)
Run the block below in a single Bash call. It runs under bash regardless of the login shell, never fails on a missing file (every field falls back to n/a), and makes no network call except the version check.
bash -s <<'PROBE'
J() { node -e 'try{const j=JSON.parse(require("fs").readFileSync(process.argv[1],"utf8"));const v=process.argv[2].split(".").reduce((o,k)=>o==null?o:o[k],j);process.stdout.write(v==null?"n/a":String(v))}catch(e){process.stdout.write("n/a")}' "$1" "$2" 2>/dev/null || echo n/a; }
has() { command -v "$1" >/dev/null 2>&1; }
P="$HOME/.claude/plugins"
# Clone root (has .git) and plugin source (has .claude-plugin/plugin.json) are
# different directories: the repo nests the plugin at commander/cowork-plugin/.
CLONE="$P/marketplaces/commander-hub"; [ -d "$CLONE" ] || CLONE="n/a"
PLUGIN_SRC="n/a"
[ "$CLONE" != "n/a" ] && [ -d "$CLONE/commander/cowork-plugin" ] && PLUGIN_SRC="$CLONE/commander/cowork-plugin"
# installed_plugins.json: the executing copy. Its installPath IS the plugin
# source (cache/<mp>/<plugin>/<version>/), with no commander/cowork-plugin wrapper.
INSTALLED_JSON="$P/installed_plugins.json"
INST=$(node -e 'try{const d=JSON.parse(require("fs").readFileSync(process.argv[1],"utf8"));const k=Object.keys(d.plugins||{}).find(k=>k.startsWith("commander@"));const e=k&&[].concat(d.plugins[k])[0];process.stdout.write(e?[k,e.version||"n/a",e.installPath||"n/a"].join("\t"):"not listed\tn/a\tn/a")}catch(e){process.stdout.write("not found or unparseable\tn/a\tn/a")}' "$INSTALLED_JSON" 2>/dev/null)
INSTALLED_KEY=$(printf '%s' "$INST" | cut -f1); INSTALLED_VERSION=$(printf '%s' "$INST" | cut -f2); INSTALLED_PATH=$(printf '%s' "$INST" | cut -f3)
[ "$PLUGIN_SRC" = "n/a" ] && [ -d "$INSTALLED_PATH" ] && PLUGIN_SRC="$INSTALLED_PATH"
# Desktop-managed (account-synced): no clone and no installed_plugins.json
# entry is HEALTHY in this mode. Update path: Desktop settings, not the CLI.
INSTALL_MODE="cli-marketplace"
if [ "$PLUGIN_SRC" = "n/a" ]; then
INSTALL_MODE="not-found"
DM=$(/bin/ls -t "$HOME/Library/Application Support/Claude/"local-agent-mode-sessions/*/*/rpm/plugin_*/.claude-plugin/plugin.json 2>/dev/null | head -50 | while read -r f; do
/usr/bin/grep -q '"name"[[:space:]]*:[[:space:]]*"commander"' "$f" 2>/dev/null && { echo "$f"; break; }; done)
[ -n "$DM" ] && { INSTALL_MODE="desktop-managed"; PLUGIN_SRC=$(dirname "$(dirname "$DM")"); }
fi
PLUGIN_VERSION=$(J "$PLUGIN_SRC/.claude-plugin/plugin.json" version)
NODE_VER=$(node --version 2>/dev/null || echo n/a)
OS_INFO=$(uname -srm 2>/dev/null || echo n/a)
GIT_SHA=n/a; [ "$CLONE" != "n/a" ] && GIT_SHA=$(git -C "$CLONE" rev-parse --short HEAD 2>/dev/null || echo n/a)
TEMP_DIRS=$(ls "$P/marketplaces/" 2>/dev/null | /usr/bin/grep '^temp_' | tr '\n' ' '); [ -z "$TEMP_DIRS" ] && TEMP_DIRS=none
# settings.json: extraKnownMarketplaces and enabledPlugins may be arrays or objects.
SETTINGS="$HOME/.claude/settings.json"
read -r HAS_MARKETPLACE ENABLED <<EOF
$(node -e 'try{const d=JSON.parse(require("fs").readFileSync(process.argv[1],"utf8"));const keys=x=>Array.isArray(x)?x.map(m=>typeof m==="string"?m:(m&&m.name)||""):Object.keys(x||{});const ep=d.enabledPlugins||{};const k=keys(ep).find(n=>/^commander(@|$)/i.test(n));const en=!k?"no":(!Array.isArray(ep)&&ep[k]===false?"disabled":"yes");process.stdout.write((keys(d.extraKnownMarketplaces).some(n=>/commander/i.test(n))?"yes":"no")+" "+en)}catch(e){process.stdout.write("n/a n/a")}' "$SETTINGS" 2>/dev/null)
EOF
MCP_SERVERS=$(node -e 'try{const s=JSON.parse(require("fs").readFileSync(process.argv[1],"utf8")).mcpServers||{};process.stdout.write(Object.keys(s).join(", ")||"none")}catch(e){process.stdout.write("n/a")}' "$PLUGIN_SRC/.mcp.json" 2>/dev/null)
SESSION_COUNT=$(ls "$HOME/.claude/sessions" 2>/dev/null | wc -l | tr -d ' ')
LATEST_SESSION=$(ls -t "$HOME/.claude/sessions" 2>/dev/null | head -1); [ -z "$LATEST_SESSION" ] && LATEST_SESSION=none
# Update freshness: shared update-check module, else the published manifest.
REMOTE_VERSION=n/a
if [ "$CLONE" != "n/a" ]; then REMOTE_VERSION=$(node "$CLONE/commander/update-check.js" --remote-only 2>/dev/null || echo n/a)
else REMOTE_VERSION=$(curl -fsSL --max-time 8 "https://raw.githubusercontent.com/KevinZai/commander/main/commander/cowork-plugin/.claude-plugin/plugin.json" 2>/dev/null | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{try{const v=JSON.parse(s).version;process.stdout.write(/^\d+\.\d+\.\d+$/.test(v)?v:"n/a")}catch(e){process.stdout.write("n/a")}})' 2>/dev/null || echo n/a); fi
CMP=$INSTALLED_VERSION; [ "$CMP" = "n/a" ] && CMP=$PLUGIN_VERSION
VERSION_STATUS="unknown — could not reach update server"
[ "$REMOTE_VERSION" != "n/a" ] && [ "$CMP" != "n/a" ] && VERSION_STATUS=$(node -e 'const [r,i]=process.argv.slice(1);const a=r.split(".").map(Number),b=i.split(".").map(Number);let gt=false;for(let n=0;n<3;n++){if((a[n]||0)>(b[n]||0)){gt=true;break}if((a[n]||0)<(b[n]||0))break}process.stdout.write(gt?"UPDATE AVAILABLE: "+i+" -> "+r:"up to date ("+i+")")' "$REMOTE_VERSION" "$CMP" 2>/dev/null || echo unknown)
# Cache dirs linger by design (no auto-clean); more than one is normal.
CACHE_ROOT="$P/cache/commander-hub/commander"
CACHE_VERSIONS=$(ls "$CACHE_ROOT" 2>/dev/null | sort -V | tr '\n' ',' | sed 's/,$//'); [ -z "$CACHE_VERSIONS" ] && CACHE_VERSIONS=none
CACHE_VERSION_COUNT=$(ls "$CACHE_ROOT" 2>/dev/null | wc -l | tr -d ' ')
INLINE_DIR="$P/data/commander-inline"; INLINE_PRESENT=no; INLINE_VERSION=n/a
[ -d "$INLINE_DIR" ] && { INLINE_PRESENT=yes; INLINE_VERSION=$(J "$INLINE_DIR/.claude-plugin/plugin.json" version); }
CATALOG_AGE_HOURS=$(node -e 'try{process.stdout.write(String(Math.floor((Date.now()-require("fs").statSync(process.argv[1]).mtimeMs)/3600000)))}catch(e){process.stdout.write("n/a")}' "$P/plugin-catalog-cache.json" 2>/dev/null)
AUTOUPDATE_FLAG=$(node -e 'try{const m=JSON.parse(require("fs").readFileSync(process.argv[1],"utf8"))["commander-hub"];process.stdout.write(m?String(!!m.autoUpdate):"not-registered")}catch(e){process.stdout.write("known_marketplaces.json not found")}' "$P/known_marketplaces.json" 2>/dev/null)
# Claude Code platform checks (absent CLI is normal for Desktop-managed installs).
CLAUDE_VER=n/a; VALIDATE=n/a; DETAILS=n/a; MOD_SUPPORT=n/a
if has claude; then
CLAUDE_VER=$(claude --version 2>/dev/null | head -1 | awk '{print $1}')
if [ "$PLUGIN_SRC" != "n/a" ]; then
VALIDATE_OUT=$(claude plugin validate --strict "$PLUGIN_SRC" 2>&1); RC=$?
[ $RC -eq 0 ] && VALIDATE=pass || VALIDATE="fail (exit $RC)"
fi
DETAILS=$(claude plugin details commander 2>&1 | head -40)
MOD_SUPPORT=$(node -e 'const v=(process.argv[1]||"").split(".").map(Number);const m=[2,1,287];let ok=true;for(let i=0;i<3;i++){if((v[i]||0)>m[i])break;if((v[i]||0)<m[i]){ok=false;break}}process.stdout.write(ok?"yes":"no (needs 2.1.287+)")' "$CLAUDE_VER" 2>/dev/null)
fi
MODS_SHIPPED=$(ls "$PLUGIN_SRC/mods" 2>/dev/null | tr '\n' ' '); [ -z "$MODS_SHIPPED" ] && MODS_SHIPPED=none
MODS_INSTALLED=$(node -e 'try{const fs=require("fs");const d=JSON.parse(fs.readFileSync(process.argv[1],"utf8")).plugins||{};const out=[];for(const k of Object.keys(d)){const e=[].concat(d[k])[0];try{if(e&&/"modules"/.test(fs.readFileSync(e.installPath+"/hooks/hooks.json","utf8")))out.push(k)}catch(x){}}process.stdout.write(out.join(" ")||"none")}catch(e){process.stdout.write("n/a")}' "$INSTALLED_JSON" 2>/dev/null)
# Codex (only when installed).
CODEX_VER="not installed"; CODEX_DOCTOR=n/a; CODEX_PLUGIN=n/a
if has codex; then
CODEX_VER=$(codex --version 2>/dev/null | head -1)
CODEX_DOCTOR_OUT=$(codex doctor --json 2>&1); RC=$?
[ $RC -eq 0 ] && CODEX_DOCTOR=pass || CODEX_DOCTOR="fail (exit $RC)"
CODEX_PLUGIN=$(codex plugin list --json 2>/dev/null | /usr/bin/grep -o '"commander[^"]*"' | head -3 | tr '\n' ' '); [ -z "$CODEX_PLUGIN" ] && CODEX_PLUGIN="commander not listed"
fi
for v in INSTALL_MODE PLUGIN_SRC PLUGIN_VERSION INSTALLED_KEY INSTALLED_VERSION INSTALLED_PATH REMOTE_VERSION VERSION_STATUS \
NODE_VER OS_INFO GIT_SHA CLONE TEMP_DIRS HAS_MARKETPLACE ENABLED MCP_SERVERS SESSION_COUNT LATEST_SESSION \
CACHE_VERSIONS CACHE_VERSION_COUNT INLINE_PRESENT INLINE_VERSION CATALOG_AGE_HOURS AUTOUPDATE_FLAG \
CLAUDE_VER VALIDATE MOD_SUPPORT MODS_SHIPPED MODS_INSTALLED CODEX_VER CODEX_DOCTOR CODEX_PLUGIN; do
printf '%s=%s\n' "$v" "${!v}"
done
echo "--- validate (last 30 lines) ---"; printf '%s\n' "${VALIDATE_OUT:-n/a}" | tail -30
echo "--- plugin details ---"; printf '%s\n' "$DETAILS"
echo "--- codex doctor --json (first 3000 chars) ---"; printf '%s' "${CODEX_DOCTOR_OUT:-n/a}" | head -c 3000; echo
PROBE
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.
- 10d ago Changed · -182 lines · +22 tokens per session 09da8c82f004
- 1mo ago First seen · 521 lines · 40 tokens per session scan D d579c36ae327
ccc-doctor is a skill published in the GitHub repository KevinZai/commander (6 stars, last pushed today), licensed MIT. It adds 62 tokens to every session and 6,080 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it D with 3 findings (downloads and executes remote code, reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
systematic-debugging
Use before proposing a permanent fix for bugs, test failures, or unexpected behavior with an unknown root cause—mandatory when symptoms are far from the cause, a prior fix failed, or the issue crosses components. Reproduce, trace backward to the root cause, fix once at the source, and add a regression test. Do not…
performance-profiler
Tier: POWERFUL Category: Engineering Domain: Performance Engineering.
advisor-strategy
Use when facing complex architectural decisions, debugging dead-ends, high-stakes code changes, security reviews, performance optimization, or when the user explicitly requests advisor consultation. Also use when context is growing large and needs management, or when a second opinion from a more capable model would…
sandbox-unblock
Diagnostic protocol to run before reporting a sandbox blocker or asking for a configuration change. Eleven checks that eliminate false positives, then a report template the person holding the settings can act on. On one measured day, six of eight reported blockers turned out to be false, all from the same handful of…
forge-diagnose
Diagnose runtime infrastructure issues — cold starts, timeouts, scaling problems, network failures. Use when asked about "infra is slow", "cold starts", "network issues", "why is this timing out", "scaling problem", "latency spikes", or "service is down".
fix-bug
Run the Fix Validation pipeline to investigate, fix, and validate a bug. Ensures deterministic pipeline execution with IssueAnalyzer, FixWriter, TestWriter (conditional), TestAudit (conditional), and FixValidator stages.