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 rules/sijeeshmiziha/visionagent/tool-designgit clone --depth 1 https://github.com/sijeeshmiziha/visionagentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/rules/sijeeshmiziha/visionagent/tool-design)<a href="https://agentmods.dev/rules/sijeeshmiziha/visionagent/tool-design"><img src="https://agentmods.dev/badge/rules/sijeeshmiziha/visionagent/tool-design.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00676 | $0.00676 |
| Opus 5 | $0.00338 | $0.00338 |
| Sonnet 5 | $0.00135 | $0.00135 |
| Haiku 4.5 | $0.00068 | $0.00068 |
Grade A, and why
tool-design 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.
How it starts
The opening of the file, as written. The whole thing — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tool Design for Agents
Self-Contained Tools
- Each tool: single clear purpose, robust to errors, unambiguous description and parameters.
- Use descriptive, typed inputs (e.g. Zod schemas); avoid overlapping or ambiguous tool contracts.
Minimal Tool Sets / Choosing the Right Tools
- Keep the tool set minimal. No bloated or overlapping tools; if a human can’t choose which tool to use, the agent can’t either.
- Agents have limited context; prefer tools that match how agents work (e.g.
search_contactsoverlist_all_contacts). - Build a few workflow-oriented tools for high-impact tasks; avoid wrapping every API endpoint.
- Consolidate: e.g.
schedule_event(find availability + schedule) instead oflist_users+list_events+create_event;search_logsinstead ofread_logs;get_customer_contextinstead of separate get/list tools. Each tool should subdivide tasks like a human would and reduce context from intermediate outputs.
Namespacing
- Use consistent prefixes or suffixes (e.g. by service:
asana_search,jira_search; by resource:asana_projects_search,asana_users_search) so agents can select the right tool. Naming can affect evaluations; try prefix vs suffix in your own evals.
Returning Meaningful Context
- Prefer high-signal, human-interpretable fields (e.g.
name,file_type) over low-level ones (uuid,mime_type). - Resolve opaque IDs to semantic labels or simple schemes (e.g. 0-indexed) to reduce retrieval hallucinations.
- Optional response_format (e.g. concise vs detailed) so the agent can request less context when IDs aren’t needed.
Token-Efficient Results
- Return minimal, relevant data; filter, aggregate, or transform before returning. Prefer summaries or slices over full dumps.
- Add pagination, range selection, filtering, or truncation with sensible defaults; cap response size where appropriate.
- Truncation messages should steer the agent (e.g. “use smaller or more targeted searches”).
- Error responses: give specific, actionable guidance and valid examples—not opaque codes or raw tracebacks.
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.
- 3d ago First seen · 49 lines · 676 tokens per session scan A 0fc66b46ae0c
tool-design is a cursor rule published in the GitHub repository sijeeshmiziha/visionagent (2 stars, last pushed 5mo ago), licensed MIT. It adds 676 tokens to every session, about $0.0034 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 cursor rules, from other repositories
cursorrules
Task queue orchestrator for AI agents. TypeScript + Hono + BullMQ + Redis.
arm_index
This file defines the installation priorities for rulesets managed by ARM.
grug-brained-dev_avoid-abstractions
Don't create abstractions until you have multiple concrete examples.
grug-brained-dev_complexity-enemy
Avoid unnecessary complexity in all forms.
grug-brained-dev_debuggable-code
Write code that is easy to debug and understand.
grug-brained-dev_simple-tools
Choose tools and libraries that are simple and well-understood.