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/arthurkatcher/grok-delegate/delegate-claudegit clone --depth 1 https://github.com/arthurkatcher/grok-delegateWhat 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 | $0.00017 | $0.00626 |
| Opus 5 | $0.00009 | $0.00313 |
| Sonnet 5 | $0.00003 | $0.00125 |
| Haiku 4.5 | $0.00002 | $0.00063 |
Grade A, and why
delegate-claude scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 51 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Delegate to local Claude Code. Models/efforts are discovered live from the installed CLI.
Launch (required)
Use exactly one run_terminal_command with background: true and timeout: 0 (or a very high timeout) so the host does not SIGTERM the companion after ~20s.
Put flags before free text. Use -- before focus so paste cannot inject flags.
Do not unquote free-form text into the shell — pass as separate argv tokens after --.
node "${GROK_PLUGIN_ROOT}/scripts/delegate-companion.mjs" claude --wait --model MODEL --effort EFFORT -- overview -- "focus text"
For complex paste, write JSON and use:
node "${GROK_PLUGIN_ROOT}/scripts/delegate-companion.mjs" claude --wait --payload-file /tmp/payload.json
Payload shape: { "positionals": ["review", "focus…"], "options": { "model": "opus" } }
- Expand the Grok task for live tools (
thinking/tool/writing). - Cancel the Grok task → companion SIGTERM kills Claude process tree.
- Do not poll status into chat.
Safety defaults
| Action | Default |
|---|---|
| overview / review / adversarial | Prefer hermetic --bare + read tools (dontAsk) when ANTHROPIC_API_KEY is set; Max/OAuth without a key auto-drops --bare (same practical mode as --trust-project) so first launch works |
| rescue | acceptEdits + file tools only (Read/Edit/Write/Glob/Grep) — no Bash |
| Shell / git CLI | Explicit --allowed-tools '…,Bash(…)' or full --yolo |
Pass explicit --bare only if you have an API key and want to force hermetic (hard-fails without a key). Prefer --cwd at a real project, not $HOME.
Auth / preflight
claude auth login or ANTHROPIC_API_KEY. Setup first if unsure.
Every action runs the same readiness check first: claude on PATH + login/API key/token. If not ready, the companion does not spawn Claude — it prints setup-style next steps and exits non-zero.
Raw arguments (agent: re-split carefully; prefer structured payload):
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.
- 3d ago First seen · 51 lines · 17 tokens per session scan A 50ff308767d8
delegate-claude is a command published in the GitHub repository arthurkatcher/grok-delegate (13 stars, last pushed 1mo ago), licensed MIT. It adds 17 tokens to every session and 626 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 2026-08-30.
Other commands, from other repositories
trellis-continue
Resume work on the current task — pick up at the right phase/step in .trellis/workflow.md.
continue
Resume work on the current task — pick up at the right phase/step in .trellis/workflow.md.
finish-work
Wrap up the current session: archive the active task (and any other completed-but-unarchived tasks the user wants to clean up) and record the session journal. Code commits are NOT done here — those happen in workflow Phase 3.4 before you invoke this command.
trellis-finish-work
Wrap up the current session: archive the active task (and any other completed-but-unarchived tasks the user wants to clean up) and record the session journal. Code commits are NOT done here — those happen in workflow Phase 3.4 before you invoke this command.
OpenSpec: Apply
Implement an approved OpenSpec change and keep tasks in sync.
diff
Compare two SKILL.md files section-by-section. Parses frontmatter and body sections independently, showing exactly what changed.