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/justinjdev/fellowship/_protocolgit 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 | $0.00000 | $0.00445 |
| Opus 5 | $0.00000 | $0.00222 |
| Sonnet 5 | $0.00000 | $0.00089 |
| Haiku 4.5 | $0.00000 | $0.00044 |
Grade A, and why
_protocol 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
Fellowship Agent Messaging Protocol
Canonical spec for how fellowship agents deliver reports and respond to lifecycle events via SendMessage.
Sync note: Agent definition files must be self-contained at runtime — an agent's markdown is injected as its system prompt with no path context, so an installed plugin's agents cannot reliably read sibling files (their working directory is the user's project, not the plugin cache). Skills are different: the Skill tool exposes the invoked skill's directory, which is why SKILL.md files can reference sibling
resources/*.mdwhile agents cannot. Each agent therefore embeds the parts of this protocol it needs inline. When you edit this spec, update the embedded copies inbalrog.md,palantir.md, andscout.md.
Sending a Report
Use SendMessage to deliver findings or status to the requesting agent:
{
"type": "message",
"recipient": "<requester>",
"content": "[markdown content]",
"summary": "[short one-line summary for logs]"
}
recipient: whoever requested the work, as provided at spawn time. SendMessage addresses agents by teammate name, not task ID.- Agents spawned by a teammate (e.g. balrog, spawned by a quest runner): the requester's teammate name from the spawn context (e.g.
"quest-auth-bug"). - Agents spawned by the fellowship lead (e.g. palantir): the lead —
"team-lead". - If no requester name was provided (standalone mode), present output directly to the user instead of using SendMessage.
- Agents spawned by a teammate (e.g. balrog, spawned by a quest runner): the requester's teammate name from the spawn context (e.g.
content: full markdown report body.summary: one-line description shown in task logs (e.g.,"balrog: 2 critical, 1 high, 0 medium, 3 low findings").
Shutdown
When you receive a shutdown request via SendMessage, respond immediately and stop:
{
"type": "shutdown_response",
"request_id": "<from the incoming message>",
"approve": true
}
Do not perform any further work after sending a shutdown response.
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 · 40 lines · 0 tokens per session scan A c2ad41bdcc77
_protocol is an agent published in the GitHub repository justinjdev/fellowship (5 stars, last pushed 18d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 445 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 2026-08-31.
Other agents, from other repositories
rapid-domain-mastery
Specialized agent for mapping unfamiliar fields from multiple sources, extracting mental models, surfacing disagreements, diagnosing misconceptions, and running oral-exam style tutoring.
architecture-designer
System architecture generation agent — designs tech stack, service boundaries, data model, API contract, infrastructure topology, and security model from SRS requirements. Produces SYSTEM-ARCHITECTURE.md, DATA-MODEL.md, API-CONTRACT.md with Architecture Decision Records for every major choice.
comparative-analyzer
Comparative analysis agent — performs side-by-side codebase comparison, architecture A/B analysis, competitive analysis, before/after delta analysis, and technology evaluation with structured comparison matrices.
api-contract-validator
API contract validation agent that ensures frontend API calls match backend endpoints, request/response types align, error codes are handled, and the API surface is consistent and well-documented.
asset-generator
AI asset generation agent — connects to image generation APIs (Nano Banana, FAL AI, Replicate), manages asset storage pipelines, generates responsive variants, and integrates assets into frontend code.
comms-assistant
Communication assistant — generates and audits README, CHANGELOG, PR descriptions, commit messages, release notes, and API documentation. Cross-references docs against actual code for accuracy.