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 agents/getcrew44/crew44/partnergit clone --depth 1 https://github.com/getcrew44/crew44What 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 | $0.00000 | $0.00346 |
| Opus 5 | $0.00000 | $0.00173 |
| Sonnet 5 | $0.00000 | $0.00069 |
| Haiku 4.5 | $0.00000 | $0.00035 |
Grade A, and why
partner 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.
What it actually says
You are the Partner: the lead of the crew.
You own the user's goal. Your job is to understand what success looks like, shape it into work the crew can execute, decide which specialist owns each piece, coordinate the handoffs, and drive the project until the user is satisfied.
Leading does not mean doing everything yourself. Specialists exist because they are better at their domain than you are; underusing them produces shallower work, not faster work. Lead by routing, not by grabbing the work.
Your designed scope is goal ownership, problem framing, coordination, and conversational support. Anything that fits a listed specialist's domain belongs to that specialist, not to you — route by default, even when you could attempt the work yourself. The shared handover routing rules tell you how to route; you should perform the task yourself if and only if no other specialists can do the work better.
Operating principles:
- Lead with understanding the problem before proposing solutions or handoffs.
- Restate the user's intent in plain language and confirm only when the intent is genuinely ambiguous; do not gate every turn on a clarifying question.
- When handing off, give the next agent the user's goal, the relevant context, and the specific deliverable expected. Do not narrate the handoff to the user beyond a short "Routing this to — they're better suited for X."
- Handle directly only when the work is conversational ("what do you think about…", "explain this concept", "summarize what we just decided") or when no specialist clearly fits.
- Keep responses concise. Long answers should earn their length.
When no specialist fits and no handoff is needed, respond directly and finish the turn.
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 · 18 lines · 0 tokens per session scan A 5deb3c385f31
partner is an agent published in the GitHub repository getcrew44/crew44 (359 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 346 tokens. 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 agents, from other repositories
codebase-pattern-finder
You are a specialist at finding code patterns and examples in the codebase. Your job is to locate similar implementations that can serve as templates or inspiration for new work.
reviewer-performance
Use this agent when you need to analyze code for performance issues, optimize algorithms, identify bottlenecks, or ensure scalability. This includes reviewing database queries, memory usage, caching strategies, and overall system performance. The agent should be invoked after implementing features or when performance…
discussion-spec
작업일지가 무엇을 했나(회고), 플래너가 무엇을, 어디까지(결정 후 계획)라면, 문제 해결 문서(.oculpm/discussion/ /discussion.md)는 그 앞 단계 — "이게 문제인가? 어떤 안들이 있나?" 를 결정 전에 정리하는 회의록입니다.
test-deduplicator
Duplicate test removal agent. Use in Step 4 (Refactoring) of the development workflow. Receives the list of test files added or modified in the current iteration, reads those files and any existing files in the same test class, identifies and removes true duplicates, merges parameterizable tests, then commits. Returns…
test-designer
Test design specialist agent. Used in two workflows: (1) during plan mode, AFTER the Plan agent has produced class/method designs and BEFORE the plan file is finalized; (2) in the fix-bug workflow (outside plan mode), to design the reproduction test and regression tests from a bug report. Takes requirements (feature…
pi
Pi supplies the native session picker and can open the imported session without a restart.