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.
git clone --depth 1 https://github.com/krishnanpandya007/support-technician-setupWrote 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/krishnanpandya007/support-technician-setup/discovering-support-tools)<a href="https://agentmods.dev/rules/krishnanpandya007/support-technician-setup/discovering-support-tools"><img src="https://agentmods.dev/badge/rules/krishnanpandya007/support-technician-setup/discovering-support-tools.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 5.1 | $0.00084 | $0.06421 |
| Opus 5 | $0.00042 | $0.03210 |
| Sonnet 5 | $0.00017 | $0.01284 |
| Haiku 4.5 | $0.00008 | $0.00642 |
Grade A, and why
discovering-support-tools scanned grade A with 1 finding 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 7d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
Every tool, regardless of `connection_type`, sits behind a single **StateAdapter** boundary — `fetch(tool, params, session) → summary`. This is the one reviewed place where read-only enforcement and **session-scoping** l How it starts
The opening of the file, as written. The whole thing — 259 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Discovering Support Tools
Overview
The runtime support agent must diagnose an end user's actual situation, not guess. It does that by calling read-only tools that fetch and check the user's live state. The user-owned state may live behind more than one read surface — SQL tables, a read API, a non-SQL store, or a pure read-only accessor — so this skill works for whatever connection types the operator enabled (default: SQL only; see "Connection types" below). It reads a product's backend and proposes, for human review:
- a tool catalog — named, fixed, parameterized read tools scoped to the authenticated end user, each tagged with its
connection_type; - per-type read-only access artifacts — for SQL, the migration SQL (a restricted read-only role, a table/column allowlist, and row-level-security policies); for any other enabled type, that type's equivalent read-only access spec (endpoint allowlist + read-scoped token for an API, read-only rules/credential for a non-SQL store, an allowlist of pure read accessors for custom);
- one per-tool source spec (the SQL query, HTTP request, non-SQL query, or named accessor the tool runs).
Core principle: the agent reads the world only through these named tools. It never gets a raw connection, never composes its own query/request/execution, never sees another user's data, and never performs a change.
Connection types — confirm what the agent reads through
Before designing anything, confirm the enabled connection types with the operator (this is decision 6 in setting-up-support-agent; the set lives in the config's connections list). State back which are enabled. Default = SQL only, in which case every output below is identical to the SQL-only behavior. The four supported types:
sql(default) — a relational database. Tools are parameterized SQL; enforcement is a read-only role + column grants + RLS / mandatory owner filter.http_api— a read API the product already exposes. Tools are single, enumeratedGET/HEADrequests against allowlisted endpoints.nosql— a non-relational store (document/key-value/etc.). Tools are scoped reads with a mandatory owner filter, over a read-only credential.custom— a pure, side-effect-free read accessor the product provides (a function that computes a verdict from data it reads). Tools name an allowlisted accessor; the model never supplies code.
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.
- 7d ago First seen · 259 lines · 6,421 tokens per session scan A e5cd5e906296
discovering-support-tools is a cursor rule published in the GitHub repository krishnanpandya007/support-technician-setup (2 stars, last pushed 2mo ago), licensed MIT. It adds 84 tokens to every session and 6,421 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). 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
python
Python best practices and patterns for modern software development with Flask and SQLite.
api-property-optionality-hygiene
Fix ApiProperty/ApiPropertyOptional optionality mismatches in DTO files; use for scheduled batch fixes or DTO edits.
django
Definitive guidelines for writing maintainable, performant, and secure Django applications, emphasizing modern best practices, clear code organization, and efficient patterns.
cursor
You are working on the checkout service. Preserve transaction integrity and auditability.
shared-libraries
Shared libraries - condition framework, inventory containers, file-backed DB, itinerary, references.
env-validation-gate
Env validation gate — every app with ≥1 required env var validates its contract at boot via Zod; raw process.env is banned outside the env module. Full pattern in .claude/skills/t2000-env-gate/.