#223 docs: CLI.md's machine-readable section omits log --jsonl and the scan evidence cap

[Anonymous]

Originally created by: fu351

Two small gaps in the same section of docs/CLI.md

1. log --jsonl is missing entirely

The Machine-readable flags table (docs/CLI.md:20-26) lists --json (for scan, doctor, policy-history) and --quiet (for scan) — but not --jsonl, even though doberman log --jsonl exists (src/doberman/cli/main.py:773-810) and has its own test file.

grep -c jsonl docs/CLI.md   # currently 0

A table whose whole purpose is to index the machine-readable surface is missing one of its four flags.

2. scan --json returns more evidence than the human view shows

• src/doberman/discovery/scan.py:162 caps each capability's evidence at 10 entries.
• src/doberman/render.py:233 truncates to evidence[:3] — but only in the human-readable risk map.

So doberman scan and doberman scan --json can legitimately show different numbers of evidence entries for the same capability. That's intended, but it's undocumented, and someone diffing the two output modes would reasonably read it as a bug.

What to do

In docs/CLI.md:

1. Add a --jsonl row to the flags table (command: log) with a short example.
2. Add one sentence to the scan --json schema section noting the JSON output may carry up to 10 evidence entries where the human view shows 3.

Done when

--jsonl appears in the flags table and examples, and the doc states the 3-vs-10 difference.

Notes

• Check both numbers against the source rather than trusting this issue — they're the kind of constant that moves.
• Docs-only, no code change. There's a companion issue covering the command table in the same file; if you want both, say so and take them together.

Comment here to claim it.

🤖 Generated with Claude Code

[Anonymous]

Ticket changed by: fu351

• status: open --> closed