is-cli-agentic
public evidence reportnpm/@anthropic-ai/claude-code

claude

offers a no-account or temporary-account path: no; the agent stops at the signup wall

markdown report
agent verdictunrankedfewer than 3 identical trials
deterministic facts54/100informational · never used for rank
rubricv1.0.0N/A never penalizes

agent trajectory

Awaiting three identical trials

Deterministic observations can identify likely friction, but they cannot license the claim that an agent passed or failed.

deterministic evidence · not the rank

Every check, its evidence, and the fix

facts 54/100

N/A is excluded from both numerator and denominator. Open a check to inspect what the probe observed.

D1

Legibility

can an agent learn this tool from its own surface?

7/15 · 47%
failD1.1machine-readable schema (`schema` / `--describe` returns JSON)0/8
evidence

no capability document; agent must parse help prose

concrete fix

Add a `schema` or `--describe` command that prints the complete command, option, input, and output schema as JSON.

passD1.2exposes an MCP surface4/4
evidence

yes

passD1.3help contains usage examples3/3
evidence

yes

D2

Structured I/O

is the output machine-parseable?

15/20 · 75%
passD2.1documents a JSON output mode10/10
evidence

yes

passD2.2no ANSI escapes when the stream is piped5/5
evidence

clean

failD2.3has a quiet mode0/5
evidence

no

concrete fix

Add `--quiet` (or `--no-progress`) so stdout can contain only the requested result.

D3

Non-blocking

does it ever hang when no human is present?

12/20 · 60%
passD3.1does not hang when stdin is a non-delivering pipe8/8
evidence

ok

passD3.2--help does not hang4/4
evidence

ok

failD3.3documents a non-interactive escape hatch0/4
evidence

no

concrete fix

Document a non-interactive mode that supplies defaults or exits with an actionable error whenever input is missing.

failD3.4cold invocation under 3s0/4
evidence

3.27s

concrete fix

Defer plugins, updates, auth, and network initialization until after argument parsing so a cold invocation returns within 3 seconds.

D4

Context economy

how many tokens does it cost to use?

7/15 · 47%
failD4.1help fits an agent context budget0/8
evidence

16890 bytes (~4223 tokens)

concrete fix

Keep top-level help concise and move detail into subcommand help or machine-readable schema output.

passD4.2does not reprint full help on every mistake7/7
evidence

0 bytes

D6

Safety rails

can an agent preview before it destroys?

0/10 · 0%
failD6.1offers a dry-run0/5
evidence

no

concrete fix

Add `--dry-run` to mutating commands; validate locally and print the exact planned changes without making network calls.

failD6.2offers an explicit confirmation bypass0/5
evidence

no

concrete fix

Add an explicit `--yes` confirmation bypass that is safe to use only when every required value is supplied.

D7

Error recovery

does a mistake produce an honest, actionable signal?

12/20 · 60%
failD7.1unknown subcommand exits non-zero0/8
evidence

exit null — SILENT FAILURE

concrete fix

Return a non-zero exit code for unknown subcommands and name the closest valid subcommand on stderr.

passD7.2unknown flag exits non-zero8/8
evidence

exit 1

passD7.3error message is terse4/4
evidence

0 bytes

D9

Cold credentials

can an agent get authenticated without a human?

12/20 · 60%
passD9.1names its credential env var in its own help surface6/6
evidence

yes

failD9.2offers a no-account or temporary-account path0/8
evidence

no; the agent stops at the signup wall

concrete fix

Offer a documented temporary, guest, or claim-later path so a cold agent can complete one useful task without signup.

passD9.3does not launch a browser auth flow in a headless context6/6
evidence

ok

claude evidence report — is-cli-agentic