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 commands/rbinar/cli-dispatch/rungit clone --depth 1 https://github.com/rbinar/cli-dispatchWhat 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 | $0.00022 | $0.01683 |
| Opus 5 | $0.00011 | $0.00842 |
| Sonnet 5 | $0.00004 | $0.00337 |
| Haiku 4.5 | $0.00002 | $0.00168 |
Grade A, and why
run 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 yesterday.
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 — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run cli-dispatch worker: $ARGUMENTS
Launch a worker via cli-dispatch-run (the deterministic no-LLM runner from 3.34.0)
and print a compact verdict summary — no babysitter, zero LLM tokens spent on orchestration.
Best for mechanical delegations with a machine-checkable --verify command.
# $ARGUMENTS is substituted TEXTUALLY into this script before bash parses it, so the
# user's quoting survives as real shell quoting — do NOT wrap this in eval (nested
# double quotes would split the prompt). Verified with a stub-binary harness.
set -- $ARGUMENTS
BACKEND="${1:-}"; PROMPT="${2:-}"; shift 2 2>/dev/null || true
case "$BACKEND" in ds|ag|cx|oc|cp) ;; *)
echo "usage: /cli-dispatch:run <backend> \"<prompt>\" [--verify '<cmd>'] [--cleanup-if-clean] [more flags]"
echo "backend: ds | ag | cx | oc | cp tip: /cli-dispatch:setup to install backends"
exit 1 ;; esac
if [ -z "$PROMPT" ]; then
echo "usage: /cli-dispatch:run <backend> \"<prompt>\" [flags]"
echo "tip: prompt is required"
exit 1
fi
RUNNER=(cli-dispatch-run)
if ! command -v cli-dispatch-run >/dev/null 2>&1; then
# Issue #150: a plugin upgrade refreshes the versioned cache dir ONLY — it never re-runs
# install.sh, so ~/.local/bin keeps whatever the last /cli-dispatch:setup installed and any
# binary introduced since (cli-dispatch-run itself, originally) is absent from PATH with
# nothing on screen explaining why. Fall back to the plugin's own copy so the documented
# zero-token flow still runs, but name the cause so the install gets fixed once.
PLUGIN_ROOT="$(bash "${CLAUDE_PLUGIN_ROOT}/scripts/resolve-plugin-root.sh" "${CLAUDE_PLUGIN_ROOT}" 2>/dev/null || true)"
if [ -n "$PLUGIN_ROOT" ] && [ -f "$PLUGIN_ROOT/scripts/cli-dispatch-run" ]; then
RUNNER=(bash "$PLUGIN_ROOT/scripts/cli-dispatch-run")
echo "cli-dispatch-run is not on PATH — using the plugin's own copy: $PLUGIN_ROOT/scripts/cli-dispatch-run"
echo "Cause: upgrading the plugin refreshes the plugin cache only; ~/.local/bin still has the wrappers your last setup installed."
echo "Fix it once with /cli-dispatch:setup — this fallback also uses whatever is in ~/.local/share, which may be stale."
else
echo "cli-dispatch-run not found on PATH, and no plugin copy is usable."
echo "If you just upgraded the plugin this is expected: the upgrade never re-runs install.sh."
echo "Fix: re-run /cli-dispatch:setup (or scripts/install.sh) to reinstall the wrappers."
echo "Fallback: use /cli-dispatch:${BACKEND}-run, or call ${BACKEND}-agent directly."
exit 1
fi
fi
RC=0
"${RUNNER[@]}" --backend "$BACKEND" --cwd "$PWD" --prompt "$PROMPT" "$@" || RC=$?
SESSIONS_ROOT="${CLI_DISPATCH_SESSIONS_DIR:-}"
[ -z "$SESSIONS_ROOT" ] && [ -d "$HOME/.cache/cli-dispatch/sessions" ] && SESSIONS_ROOT="$HOME/.cache/cli-dispatch/sessions"
[ -z "$SESSIONS_ROOT" ] && SESSIONS_ROOT="$HOME/.cache/claude-ds/sessions"
# Newest session dir that actually carries a verdict.json (this run's, unless none was written).
SESSION_DIR=""
for d in $(ls -dt "$SESSIONS_ROOT"/*/ 2>/dev/null | head -5); do
[ -f "$d/verdict.json" ] && { SESSION_DIR="${d%/}"; break; }
done
if [ -n "$SESSION_DIR" ]; then
node -e '
const {readFileSync} = require("fs");
const exit = process.argv[2];
let v;
try { v = JSON.parse(readFileSync(process.argv[1], "utf8")) }
catch (e) { console.log("exit: " + exit + " (verdict.json unreadable: " + e.message + ")"); process.exit(0) }
if (v.error) { console.log("exit: " + exit + " verdict error: " + v.error); process.exit(0) }
const verify = v.verify ? (v.verify.exitCode === 0 ? "pass" : "FAIL (exit " + v.verify.exitCode + ")") : "n/a";
const diff = v.diffstat || (v.changedFiles ? v.changedFiles.length + " file(s)" : "n/a");
console.log("exit: " + exit + " session: " + (v.sessionId || "?") + " state: " + (v.state || "?") + " verify: " + verify);
console.log("diff: " + String(diff).trim());
if (v.stranded) console.log("STRANDED changes in worktree: " + v.worktree);
console.log("patch: " + (v.diffPatchPath || "n/a"));
' "$SESSION_DIR/verdict.json" "$RC"
else
echo "exit: $RC (no verdict.json found)"
fi
exit $RC
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.
- yesterday First seen · 108 lines · 22 tokens per session scan A fa035806eca0
run is a command published in the GitHub repository rbinar/cli-dispatch (5 stars, last pushed 13d ago), licensed MIT. It adds 22 tokens to every session and 1,683 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 commands, from other repositories
strict
Enable strict RIPER protocol enforcement.
research
Enter RESEARCH mode for information gathering.
status
Show the current status of the Claude Code Router server.
brainstorm
多模型并行对同一问题各给独立意见(发散式 brainstorming),主 Claude 综合分歧、共识、独到见解。各模型彼此看不到对方答案,避免回声室效应。.
language
Toggle the HUD label language between English and 中文 (edits /.claude/plugins/balance-hud/config.json).
benchmark
Run the Permafrost cache benchmark (offline emulator, no API key).