create-agent-context

A guided process for producing the JSON profile that Vector uses to describe a saved LLM agent. Vector is a service for testing AI agents with red-team attacks.

In plain words
What is it for?
It is for interviewing the user, creating an agent slug, and generating the five-field AgentContext JSON to paste into Vector.
Why use it?
It turns an agent description into the specific fields and format required by Vector's New Agent form.

Skill for Claude CodeCodex

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 skills/pharosone/vector-plugin/create-agent-context
Any agent
npx skills add pharosone/vector-plugin --skill create-agent-context
Clone the repo
git clone --depth 1 https://github.com/pharosone/vector-plugin

Made for: Claude Code, Codex.

Per session 62 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,177 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 $0.00062 $0.01177
Opus 5 $0.00031 $0.00589
Sonnet 5 $0.00012 $0.00235
Haiku 4.5 $0.00006 $0.00118

Measured yesterday against content hash 7da3a1cef032, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

create-agent-context 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.

skills/create-agent-context/SKILL.md · 65 lines

How it starts

The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Goal

Generate an AgentContext payload for a Vector saved agent. The user will paste it into the New Agent page on the cabinet.

Context

  • Cabinet (default SaaS deployment): https://vector.pharosone.ai. REST API base: https://vector-api.pharosone.ai. If the user is on a different deployment, ask them for the URLs first.
  • Live schema (CORS-open, no auth): https://vector-api.pharosone.ai/api/v1/public/agent-context-schema.json. Fetch this once at the start if your editor has web access — every field name used below must appear under components.schemas.AgentContext.properties. If it doesn't, the schema has been updated and you should trust the fetched schema, not this skill.

The page has TWO inputs (you must produce TWO things)

  1. A plain text input labelled "Name" above the JSON editor — that's the agent's slug (e.g. support-bot-v3). It is NOT inside the JSON.
  2. The JSON editor — its content is exactly an AgentContext object (the 5-field schema below). No { name, description, agent_context: {...} } wrapper around it.

AgentContext — use ONLY these 5 fields, with these EXACT names

Required:

  • description (string, 1–8000 chars) — Free-form prose: what the agent does, who it serves, what makes it risky, and (if relevant) which tools / data / external systems it can touch — written as prose, not a structured list. This is the single most important field; the planner reads it to choose attack categories. Aim for 3–6 sentences.

Optional, but each one sharpens a different attack class:

  • name (string, ≤200) — human-readable display label shown in cabinet/reports (NOT the slug).
  • language_hints (string[]) — Always set to exactly ["en"]. This deployment supports only English-language agents.
  • restrictions (string[]) — short imperative lines listing things the agent must NEVER do. Drives rule-evasion / policy-bypass attacks. 3–8 entries is a good range.
  • system_prompt_excerpt (string, target ≤500 chars; hard limit 4000) — a short, tight snippet of the user's real system prompt: 2–4 sentences picked specifically because they contain policy verbs ("do not", "never", "always", "must", "refuse", "only", "before") or hard limits ("up to $500", "max 3 retries"). The planner uses these as attack anchors — a multi-paragraph dump adds noise, a focused snippet sharpens attacks. Paraphrase only if the real prompt is confidential.

Read the full file on GitHub · 65 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. yesterday First seen · 65 lines · 62 tokens per session scan A 7da3a1cef032

Subscribe to this mod's changes

create-agent-context is a skill published in the GitHub repository pharosone/vector-plugin (1 stars, last pushed 3mo ago), licensed MIT. It adds 62 tokens to every session and 1,177 once invoked, about $0.0003 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.

Related

Other skills, from other repositories