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/justinjdev/fellowship/guidegit clone --depth 1 https://github.com/justinjdev/fellowshipWhat 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 | $0.00027 | $0.01331 |
| Opus 5 | $0.00014 | $0.00665 |
| Sonnet 5 | $0.00005 | $0.00266 |
| Haiku 4.5 | $0.00003 | $0.00133 |
Grade A, and why
guide 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Fellowship Guide — Learn by Doing
Overview
This guide teaches fellowship by running a real task on the user's codebase. It does NOT explain concepts upfront — it demonstrates them, then names them afterward.
Do not mention gates, lembas, tome, worktree, balrog, warden, council, palantir, herald, eagles, errands, companies, or scouts until Act 3 (Graduation). Use plain language throughout Acts 1–2.
Act 1: Pitch
Present this to the user:
Fellowship helps Claude work through complex tasks more carefully — researching before planning, planning before coding. Each step has a checkpoint where you review the work before moving on.
Let's try it on something real in your codebase.
Then ask using AskUserQuestion:
What would you like to work on?
Pick something real — a bug to fix, a feature to add, or a refactor to do. It should be non-trivial (more than a one-line change) but completable in a single session.
Wait for their response. This is the task description used for the rest of the guide.
Act 2: Guided Quest
Setup
- Determine the current branch:
git branch --show-current - Create a new branch:
git checkout -b guide/<slug>where<slug>is a short kebab-case summary of the task (e.g.,guide/fix-auth-redirect) - Read the project's CLAUDE.md (if it exists) to understand conventions and reference files. Do not present this to the user — absorb it silently.
Stage 1: Research
Goal: Understand the relevant parts of the codebase before making a plan.
- Use the Explore agent (Agent tool with subagent_type=Explore, passing
model: "haiku"— ormodels.explorefrom fellowship config if set) to find files related to the task. Focus on:- Files that will likely need modification
- Files that define patterns to follow
- Test files for the affected area
- Types, configs, or dependencies that constrain the work
- If CLAUDE.md lists reference files for the relevant area, read them and note the patterns (naming, error handling, structure, data flow). Do not present the raw pattern analysis — internalize it.
- Present a structured research summary to the user:
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 · 144 lines · 27 tokens per session scan A ad0f5511a071
guide is a command published in the GitHub repository justinjdev/fellowship (5 stars, last pushed 18d ago), licensed Apache-2.0. It adds 27 tokens to every session and 1,331 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
auto-optimize
Self-improving agent optimization — generates challenger variants of any agent/command, benchmarks against baseline, promotes winners, logs learnings to instincts. Inspired by Karpathy's autoresearch pattern.
learn-mode
Interactive code tutor — breaks down codebase logic, explains complexities, translates technical concepts for the user. Ideal after /btw commands. Teaches the WHY behind the code, not just the WHAT.
session-validate
End-of-session self-training — captures session metrics, extracts patterns via metaclaw-learner, updates instincts, and generates optimization hypotheses for the next run.
hire-debrief
Run the post-loop decision debrief — facilitator script for the 45-60 minute session. Surface inter-rater drift, anchor every claim to the rubric, name bias patterns out loud, render the hire-or-no-hire decision per the pre-committed rule, and capture learning for the next loop. Structured scores submitted BEFORE…
handover
End-of-session handover with wisdom capture. Writes status + reflections to docs/ops/, Starlight Vaults, and optionally GitHub.
lavra-checkpoint
Save session progress by filing beads, capturing knowledge, and syncing state.