Agent DX CLI Scale
Use this skill to evaluate any CLI against the principles of agent-first design. Score each axis from 0–3, then sum for a total between 0–21.
Human DX optimizes for discoverability and forgiveness.
Agent DX optimizes for predictability and defense-in-depth.
— You Need to Rewrite Your CLI for AI Agents
Scoring Axes
1. Machine-Readable Output
Can an agent parse the CLI's output without heuristics?
| Score |
Criteria |
| 0 |
Human-only output (tables, color codes, prose). No structured format available. |
| 1 |
--output json or equivalent exists but is incomplete or inconsistent across commands. |
| 2 |
Consistent JSON output across all commands. Errors also return structured JSON. |
| 3 |
NDJSON streaming for paginated results. Structured output is the default in non-TTY (piped) contexts. |
2. Raw Payload Input
Can an agent send the full API payload without translation through bespoke flags?
| Score |
Criteria |
| 0 |
Only bespoke flags. No way to pass structured input. |
| 1 |
Accepts --json or stdin JSON for some commands, but most require flags. |
| 2 |
All mutating commands accept a raw JSON payload that maps directly to the underlying API schema. |
| 3 |
Raw payload is first-class alongside convenience flags. The agent can use the API schema as documentation with zero translation loss. |
3. Schema Introspection
Can an agent discover what the CLI accepts at runtime without pre-stuffed documentation?
| Score |
Criteria |
| 0 |
Only --help text. No machine-readable schema. |
| 1 |
--help --json or a describe command for some surfaces, but incomplete. |
| 2 |
Full schema introspection for all commands — params, types, required fields — as JSON. |
| 3 |
Live, runtime-resolved schemas (e.g., from a discovery document) that always reflect the current API version. Includes scopes, enums, and nested types. |
4. Context Window Discipline
Does the CLI help agents control response size to protect their context window?
| Score |
Criteria |
| 0 |
Returns full API responses with no way to limit fields or paginate. |
| 1 |
Supports --fields or field masks on some commands. |
| 2 |
Field masks on all read commands. Pagination with --page-all or equivalent. |
| 3 |
Streaming pagination (NDJSON per page). Explicit guidance in context/skill files on field mask usage. The CLI actively protects the agent from token waste. |
5. Input Hardening
Does the CLI defend against the specific ways agents fail (hallucinations, not typos)?
| Score |
Criteria |
| 0 |
No input validation beyond basic type checks. |
| 1 |
Validates some inputs, but does not cover agent-specific hallucination patterns (path traversals, embedded query params, double encoding). |
| 2 |
Rejects control characters, path traversals (../), percent-encoded segments (%2e), and embedded query params (?, #) in resource IDs. |
| 3 |
Comprehensive hardening: all of the above, plus output path sandboxing to CWD, HTTP-layer percent-encoding, and an explicit security posture — "The agent is not a trusted operator." |
6. Safety Rails
Can agents validate before acting, and are responses sanitized against prompt injection?
| Score |
Criteria |
| 0 |
No dry-run mode. No response sanitization. |
| 1 |
--dry-run exists for some mutating commands. |
| 2 |
--dry-run for all mutating commands. Agent can validate requests without side effects. |
| 3 |
Dry-run plus response sanitization (e.g., via Model Armor) to defend against prompt injection embedded in API data. The full request→response loop is defended. |
7. Agent Knowledge Packaging
Does the CLI ship knowledge in formats agents can consume at conversation start?
| Score |
Criteria |
| 0 |
Only --help and a docs site. No agent-specific context files. |
| 1 |
A CONTEXT.md or AGENTS.md with basic usage guidance. |
| 2 |
Structured skill files (YAML frontmatter + Markdown) covering per-command or per-API-surface workflows and invariants. |
| 3 |
Comprehensive skill library encoding agent-specific guardrails ("always use --dry-run", "always use --fields"). Skills are versioned, discoverable, and follow a standard like OpenClaw. |
Interpreting the Total
| Range |
Rating |
Description |
| 0–5 |
Human-only |
Built for humans. Agents will struggle with parsing, hallucinate inputs, and lack safety rails. |
| 6–10 |
Agent-tolerant |
Agents can use it, but they'll waste tokens, make avoidable errors, and require heavy prompt engineering to compensate. |
| 11–15 |
Agent-ready |
Solid agent support. Structured I/O, input validation, and some introspection. A few gaps remain. |
| 16–21 |
Agent-first |
Purpose-built for agents. Full schema introspection, comprehensive input hardening, safety rails, and packaged agent knowledge. |
Bonus: Multi-Surface Readiness
Not scored, but note whether the CLI exposes multiple agent surfaces from the same binary:
1---2name: agent-dx-cli-scale3description: A scoring scale for evaluating how well a CLI is designed for AI agents, based on the "Rewrite Your CLI for AI Agents" principles.4---56# Agent DX CLI Scale78Use this skill to **evaluate any CLI** against the principles of agent-first design. Score each axis from 0–3, then sum for a total between 0–21.910> Human DX optimizes for discoverability and forgiveness.11> Agent DX optimizes for predictability and defense-in-depth.12> — [You Need to Rewrite Your CLI for AI Agents](/posts/rewrite-your-cli-for-ai-agents)1314---1516## Scoring Axes1718### 1. Machine-Readable Output1920Can an agent parse the CLI's output without heuristics?2122| Score | Criteria |23| ----- | ----------------------------------------------------------------------------------------------------- |24| 0 | Human-only output (tables, color codes, prose). No structured format available. |25| 1 | `--output json` or equivalent exists but is incomplete or inconsistent across commands. |26| 2 | Consistent JSON output across all commands. Errors also return structured JSON. |27| 3 | NDJSON streaming for paginated results. Structured output is the default in non-TTY (piped) contexts. |2829### 2. Raw Payload Input3031Can an agent send the full API payload without translation through bespoke flags?3233| Score | Criteria |34| ----- | ------------------------------------------------------------------------------------------------------------------------------------- |35| 0 | Only bespoke flags. No way to pass structured input. |36| 1 | Accepts `--json` or stdin JSON for some commands, but most require flags. |37| 2 | All mutating commands accept a raw JSON payload that maps directly to the underlying API schema. |38| 3 | Raw payload is first-class alongside convenience flags. The agent can use the API schema as documentation with zero translation loss. |3940### 3. Schema Introspection4142Can an agent discover what the CLI accepts at runtime without pre-stuffed documentation?4344| Score | Criteria |45| ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |46| 0 | Only `--help` text. No machine-readable schema. |47| 1 | `--help --json` or a `describe` command for some surfaces, but incomplete. |48| 2 | Full schema introspection for all commands — params, types, required fields — as JSON. |49| 3 | Live, runtime-resolved schemas (e.g., from a discovery document) that always reflect the current API version. Includes scopes, enums, and nested types. |5051### 4. Context Window Discipline5253Does the CLI help agents control response size to protect their context window?5455| Score | Criteria |56| ----- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |57| 0 | Returns full API responses with no way to limit fields or paginate. |58| 1 | Supports `--fields` or field masks on some commands. |59| 2 | Field masks on all read commands. Pagination with `--page-all` or equivalent. |60| 3 | Streaming pagination (NDJSON per page). Explicit guidance in context/skill files on field mask usage. The CLI actively protects the agent from token waste. |6162### 5. Input Hardening6364Does the CLI defend against the specific ways agents fail (hallucinations, not typos)?6566| Score | Criteria |67| ----- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |68| 0 | No input validation beyond basic type checks. |69| 1 | Validates some inputs, but does not cover agent-specific hallucination patterns (path traversals, embedded query params, double encoding). |70| 2 | Rejects control characters, path traversals (`../`), percent-encoded segments (`%2e`), and embedded query params (`?`, `#`) in resource IDs. |71| 3 | Comprehensive hardening: all of the above, plus output path sandboxing to CWD, HTTP-layer percent-encoding, and an explicit security posture — _"The agent is not a trusted operator."_ |7273### 6. Safety Rails7475Can agents validate before acting, and are responses sanitized against prompt injection?7677| Score | Criteria |78| ----- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |79| 0 | No dry-run mode. No response sanitization. |80| 1 | `--dry-run` exists for some mutating commands. |81| 2 | `--dry-run` for all mutating commands. Agent can validate requests without side effects. |82| 3 | Dry-run plus response sanitization (e.g., via Model Armor) to defend against prompt injection embedded in API data. The full request→response loop is defended. |8384### 7. Agent Knowledge Packaging8586Does the CLI ship knowledge in formats agents can consume at conversation start?8788| Score | Criteria |89| ----- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |90| 0 | Only `--help` and a docs site. No agent-specific context files. |91| 1 | A `CONTEXT.md` or `AGENTS.md` with basic usage guidance. |92| 2 | Structured skill files (YAML frontmatter + Markdown) covering per-command or per-API-surface workflows and invariants. |93| 3 | Comprehensive skill library encoding agent-specific guardrails (_"always use --dry-run"_, _"always use --fields"_). Skills are versioned, discoverable, and follow a standard like OpenClaw. |9495---9697## Interpreting the Total9899| Range | Rating | Description |100| ----- | ------------------ | ------------------------------------------------------------------------------------------------------------------------------- |101| 0–5 | **Human-only** | Built for humans. Agents will struggle with parsing, hallucinate inputs, and lack safety rails. |102| 6–10 | **Agent-tolerant** | Agents can use it, but they'll waste tokens, make avoidable errors, and require heavy prompt engineering to compensate. |103| 11–15 | **Agent-ready** | Solid agent support. Structured I/O, input validation, and some introspection. A few gaps remain. |104| 16–21 | **Agent-first** | Purpose-built for agents. Full schema introspection, comprehensive input hardening, safety rails, and packaged agent knowledge. |105106---107108## Bonus: Multi-Surface Readiness109110Not scored, but note whether the CLI exposes multiple agent surfaces from the same binary:111112- [ ] **MCP (stdio JSON-RPC)** — typed tool invocation, no shell escaping113- [ ] **Extension / plugin install** — agent treats the CLI as a native capability114- [ ] **Headless auth** — env vars for tokens/credentials, no browser redirect required
peteedoo