CLI-Anything Bundle #6 — Format Recon Pack

Four agent-native CLI skills that ask the file questions before you parse, import, or summarize anything: jsonprobe (what shape is this JSON, and is it even true JSON), csvpeek (what is really inside this CSV), mdskim (what does this markdown actually contain), encpeek (which bytes are lying about being text). Python 3.8+, zero dependencies, Apache 2.0.

Safety contract: all four skills are strictly read-only over everything they inspect. No writes, no deletes, no moves, no conversions, no permission changes, no network. Delimiters, headers, and encodings are reported together with how they were decided — never guessed silently. Ragged rows, blank lines, and NUL bytes are counted, never dropped silently. jsonprobe demands valid UTF-8 and names the first bad byte instead of decoding through mojibake.

1. Unzip

unzip CLI-Anything_Bundle6_FormatRecon_v1.0.zip

2. Use with your AI agent

Copy any skill folder (the one containing its SKILL.md) into your agent's skills directory — for OpenClaw that is ~/.openclaw/skills/:

cp -r jsonprobe ~/.openclaw/skills/
cp -r csvpeek ~/.openclaw/skills/
cp -r mdskim ~/.openclaw/skills/
cp -r encpeek ~/.openclaw/skills/

3. Or run as a plain CLI

python3 jsonprobe/agent-harness/cli_anything/jsonprobe/__main__.py summary config.json
python3 csvpeek/agent-harness/cli_anything/csvpeek/__main__.py columns data.csv
python3 mdskim/agent-harness/cli_anything/mdskim/__main__.py headings notes.md
python3 encpeek/agent-harness/cli_anything/encpeek/__main__.py file data.txt

Every command accepts --json (before or after the subcommand) and reads - as stdin where a PATH is taken. Duplicate keys, ragged rows, and unterminated fences are reported honestly — JSON forbids them, real files have them.

4. Verify before you trust (122 checks)

python3 jsonprobe/agent-harness/cli_anything/jsonprobe/tests/test_core.py  # 29 checks
python3 csvpeek/agent-harness/cli_anything/csvpeek/tests/test_core.py      # 24 checks
python3 mdskim/agent-harness/cli_anything/mdskim/tests/test_core.py        # 35 checks
python3 encpeek/agent-harness/cli_anything/encpeek/tests/test_core.py      # 34 checks

Each battery prints PASS/FAIL per check and exits non-zero on any failure. The batteries also live-fire each skill's real CLI end-to-end (verified 2026-10-18: 122 checks, 0 failed).

5. Command surface

skillcommands
jsonprobesummary PATH · paths PATH [--max N] · dupes PATH · validate PATH
csvpeeksummary PATH · columns PATH · sample PATH [--rows N] · ragged PATH
mdskimskim PATH · headings PATH · links PATH [--check-files] · scan PATH
encpeekfile PATH · lines PATH · check PATH [--bom ok|require|reject] [--require-final-newline]

6. The family (all on the free rail)

License: Apache 2.0 — yours to run, modify, and ship anywhere, including inside your own products. Support: [email protected].

Built on SoulTech AI by Kiwann Williams. All Is One. One Is All.