agent-boundaries

A guide for deciding which airline-travel agent owns a request or operation. It distinguishes a single operation, such as booking, from a multi-step workflow, such as searching, pricing, booking, and ticketing.

In plain words
What is it for?
Use it to assign travel requests to the right agent, resolve responsibility boundaries, and decide whether a workflow needs multi-agent orchestration.
Why use it?
It prevents overlapping agents from handling the same responsibility and helps route requests to the correct part of the system. It also clarifies when an orchestrator should coordinate several agents.

Agent

Install

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.

agentmods
npx agentmods add agents/telivityai/otaip/agent-boundaries
Clone the repo
git clone --depth 1 https://github.com/TelivityAI/otaip
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,256 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.03256
Opus 5 $0.00000 $0.01628
Sonnet 5 $0.00000 $0.00651
Haiku 4.5 $0.00000 $0.00326

Measured 3d ago against content hash 5e21c9894dae, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-boundaries 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.

docs/agents/agent-boundaries.md · 142 lines

How it starts

The opening of the file, as written. The whole thing — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Boundaries

Many OTAIP stages contain sibling agents with adjacent responsibilities. This guide states which agent owns which responsibility — the intent, field, or operation that assigns a request to exactly one agent — derived from the agent contracts.

Use it to assign a request to the right agent: find the relevant agents, and the Boundary line is the deciding signal. Each agent is identified by its readonly id (the binding identifier).

Scope note: every boundary here is derived from the agents' own contracts in packages/. When a contract is genuinely ambiguous it is called out inline.


Orchestration vs. single-purpose — Agent 9.1

This one ownership rule covers the most agents. 9.1 Orchestrator owns a request only when it requires a multi-step, multi-agent workflow — its contract defines exactly these pipelines: search-to-price, book-to-ticket, full booking, exchange flow, refund flow (with conditional/parallel steps and error handling). A single concrete operation belongs to its specific agent, never 9.1.

Belongs to 9.1 (multi-step) Belongs to the specific agent (single op)
"search → price → book → ticket this end to end" one availability query → 1.1
"reconcile, then refund, then update loyalty" one BSP reconciliation → 7.1
"run the exchange flow (assess + reissue + collect)" one pre-ticket audit → 6.2
any chain that hands output between ≥2 agents profile read / apply-to-PNR → 8.1
mid-office PNR QC / deadline scan → 8.3

Note: an agent that internally composes others is not automatically 9.1. For example 5.5 Self-Service Rebooking composes 1.1 + 5.1, but it is a single named capability ("present priced rebooking alternatives") — still 5.5, not 9.1. The 9.1 signal is the user asking for a multi-stage pipeline, not an agent happening to call others.


Reference data (0.x)

Pair A B Boundary
0.1 ↔ 0.6 0.1 Airport/City Code Resolver — IATA/ICAO airport & city codes, multi-airport cities 0.6 Currency & Tax Code Resolver — ISO 4217 currency + IATA tax/surcharge codes A resolves a place code; B resolves a money/tax code
0.3 ↔ 0.6 0.3 Fare Basis Code Decoder — decodes fare basis into cabin, restrictions, advance-purchase, penalties 0.6 Currency & Tax Code Resolver A decodes a fare basis; B resolves a currency/tax code
0.3 ↔ 3.3 0.3 Fare Basis Code Decoder 3.3 PNR Validation — pre-ticketing PNR checks A decodes a code; B validates a PNR

Read the full file on GitHub · 142 lines

Changes

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.

  1. 3d ago First seen · 142 lines · 0 tokens per session scan A 5e21c9894dae

Subscribe to this mod's changes

agent-boundaries is an agent published in the GitHub repository TelivityAI/otaip (40 stars, last pushed 9d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 3,256 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.