skill-testing

Testing rules for agent skills based on file structure, real scripts, and tool outputs. They distinguish tests that check actual behavior from tests that merely freeze the wording of documentation.

In plain words
What is it for?
They help maintain skill metadata, test skill scripts with real inputs and outputs, validate MCP response fields, and decide when behavior needs an agent evaluation.
Why use it?
They prevent misleading tests that pass because text was copied rather than because the skill works. They also define the checks expected for skills, scripts, and MCP tools.

Cursor rule for Cursor

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 rules/granoflow/granoflow-mcp-server/skill-testing
Clone the repo
git clone --depth 1 https://github.com/granoflow/granoflow-mcp-server

Made for: Cursor.

Per session 383 This file is loaded in full into every session.
When invoked 383 The same file — it is already loaded in full.
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 $0.00383 $0.00383
Opus 5 $0.00192 $0.00192
Sonnet 5 $0.00077 $0.00077
Haiku 4.5 $0.00038 $0.00038

Measured yesterday against content hash ef98a506f916, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

skill-testing 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.

.cursor/rules/skill-testing.mdc · 38 lines

What it actually says

Skill Testing

Follow popular Agent Skills practice (Anthropic / agentskills.io style):

Allowed in CI

  • Structure: SKILL.md exists; YAML frontmatter has non-empty name and description; name matches the skill directory; description length ≤ 1024. Covered by tests/skills-structure.test.ts.
  • Scripts: unit/integration tests for real skills/*/scripts I/O and error codes under tests/python/.
  • MCP surface: skill tools assert ok, data.path, non-empty skill (optionally starts with ---), and optional references[].referenceId presence — not body phrases.

Forbidden

Do not add or restore tests that treat skill markdown as a snapshot:

// FORBIDDEN — prose / gate-copy contract locks
const skill = readFileSync("skills/granoflow-…/SKILL.md", "utf8");
expect(skill).toContain("some workflow sentence");
expect(reference).toMatch(/hard.gate.phrase/i);
expect(JSON.stringify(skillToolResult)).toContain("activation phrase");

That pattern does not prove agent behavior; it only freezes wording and creates false confidence.

Behavior correctness

Use LLM evals (trigger + functional expectations) when validating that agents follow a skill. This repo does not require eval harness in CI today. Do not substitute evals with markdown toContain locks.

Planning / ask-mode

Do not invent natural-language Excel “test cases” for prose-only skill edits. For those changes: files + intent → edit → npm run check. See change-quality-gate.mdc.

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 · 38 lines · 383 tokens per session scan A ef98a506f916

Subscribe to this mod's changes

skill-testing is a cursor rule published in the GitHub repository granoflow/granoflow-mcp-server (0 stars, last pushed 1mo ago), licensed MIT. It adds 383 tokens to every session, about $0.0019 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.