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/shep-ai/shep/supervisor-agentgit clone --depth 1 https://github.com/shep-ai/shepWhat 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 | $0.01687 |
| Opus 5 | $0.00000 | $0.00843 |
| Sonnet 5 | $0.00000 | $0.00337 |
| Haiku 4.5 | $0.00000 | $0.00169 |
Grade A, and why
supervisor-agent 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 2d 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 — 221 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Supervisor Agent
Purpose
The Supervisor Agent is responsible for evaluating agent collaboration events and deciding whether Shep should approve, reject, escalate, or advise on the next step.
It observes events from the collaboration fabric and helps enforce user-delegated policy for:
- approval gates
- agent questions
- inter-agent messages
The supervisor is part of spec 093 and is gated behind the collaboration feature flag. With the flag off, the collaboration and supervision surface stays disabled.
The long-form architecture lives in ../architecture/supervision.md. This page documents the per-agent identity, prompt source, output schema, flow, and operational notes.
Agent Identity
| Property | Value |
|---|---|
| Agent | supervisor |
| Lane | agents |
| Registration | ISupervisorAgent via the agent DI rail |
| Integration Point | EvaluateSupervisorDecisionUseCase + packages/core/src/infrastructure/services/agents/supervisor-agent/supervisor-agent-worker.ts |
| Spec | specs/093-agent-collaboration-supervision/spec.yaml |
| Flag Gate | collaboration |
The core evaluation entry point is EvaluateSupervisorDecisionUseCase, which:
- checks the
collaborationfeature flag - resolves the effective supervisor policy
- calls
ISupervisorAgent.evaluate(...) - persists the immutable
SupervisorDecision - mirrors the decision into
activity_logwithactor_id = "supervisor:<id>"
The worker entry point is supervisor-agent-worker.ts. It owns a supervisor AgentRun, heartbeats while active, evaluates submitted events, and reaps idle workers after the configured TTL.
System Prompt
The evaluator behavior is defined in the supervisor prompt source and should be treated as the source of truth.
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.
- 2d ago First seen · 221 lines · 0 tokens per session scan A 134033eb532a
supervisor-agent is an agent published in the GitHub repository shep-ai/shep (247 stars, last pushed 4d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,687 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
builder
Turn shot-plan.json into one renderable HyperFrames composition (compositions/index.html). Everything stays in the HF ecosystem — HTML is the source of truth; a single paused GSAP timeline carries all motion; the engine seeks it. Category-specific build rules live in categories/ /module.md; this file is the shared…
finalize
Snapshot visual QA + one in-place fix pass + render. Dispatched only when Step 6 lint/inspect reports issues, or to do the final render.
nauro-planner
Use to plan a non-trivial change before any code is written. Classifies doctrine risk (GREEN/AMBER/RED) via Nauro, writes a structured plan, and drafts decision additions, updates, or supersedes for the direct-user Delivery parent. Returns a plan and never writes project truth or edits files.
nauro-tech-lead
Use to set or maintain project direction. Reads Nauro decisions, session transcripts, and PR diffs; judges architectural choices against active doctrine; and drafts complete decision additions, updates, or supersedes for the direct-user Delivery parent. Never writes project truth.
ampl-explainer
Optimization expert that explains AMPL models: variable-constraint linkage, logical vs algebraic constraints, problem classification, and complexity analysis. Use when the user asks why a model behaves a certain way, wants a walkthrough, needs logic mapped to algebra, or has unclear modeling decisions. Readonly …
debugger
Debugging specialist for errors and test failures. Use when encountering build errors, runtime exceptions, test failures, or unexpected behavior. Invoke with /debugger to investigate issues.