tidy-skill

skills CLI verification log

Date: 2026-08-05 (Codex); 2026-08-07 (claude-code/cursor/pi added); 2026-08-08 re-verified — author machine, Windows
CLI: npx skills (vercel-labs/skills style package)
Source: https://github.com/Phoenix0531-sudo/tidy-skill.git

2026-08-08 re-verification (network recovered)

After github.com connectivity was restored, re-ran discovery + all four target installs in a fresh empty %TEMP% folder. All --copy mode.

Step Command Result
Discover npx.cmd skills add Phoenix0531-sudo/tidy-skill --list -y Found 1 skill: tidy-skill (exit 0)
Install codex ... --skill tidy-skill -a codex -y --copy Installed 1 skill (exit 0)
Install claude-code ... -a claude-code ... Installed 1 skill (exit 0)
Install cursor ... -a cursor ... Installed 1 skill (exit 0)
Install pi ... -a pi ... Installed 1 skill (exit 0)

Landing trees confirmed (one SKILL.md each, full tree: agents/ commands/ examples/ hooks/ references/ scripts/ templates/):

Frontmatter of the installed SKILL.md matches the current repo name/description. This closes the transient network-block from earlier runs.

List

Command:

npx.cmd skills add Phoenix0531-sudo/tidy-skill --list -y

Result:

Found 1 skill
tidy-skill

Project install (Codex agent, copy mode)

Command:

npx.cmd skills add Phoenix0531-sudo/tidy-skill --skill tidy-skill -a codex -y --copy

Working directory: temporary empty folder under %TEMP%.

Result:

Installed 1 skill
✓ tidy-skill (copied)
→ <temp>/.agents/skills/tidy-skill

Observed tree (abbreviated):

.agents/skills/tidy-skill/
├─ SKILL.md
├─ agents/openai.yaml
├─ commands/
├─ hooks/
├─ scripts/
├─ templates/
├─ references/
└─ examples/

Also wrote skills-lock.json in the project root.

Security panel (CLI-printed, third-party scanners): Gen=Safe, Socket=0 alerts, Snyk=Low Risk.
Details URL printed by CLI: https://skills.sh/Phoenix0531-sudo/tidy-skill

Windows note

PowerShell may block npx.ps1 under restricted execution policy. Use:

npx.cmd skills add Phoenix0531-sudo/tidy-skill --skill tidy-skill

Other agent targets (2026-08-07)

Re-verified after github.com connectivity was restored. Same command shape, -a <agent> swapped. All --copy mode in a temporary empty %TEMP% folder.

-a value Landing tree Result
claude-code .claude/skills/tidy-skill/ ✓ tidy-skill (copied)
cursor .agents/skills/tidy-skill/ ✓ tidy-skill (copied)
pi .pi/skills/tidy-skill/ ✓ tidy-skill (copied)

Adapter quirk worth recording: claude is not a valid agent name. The CLI rejects it with Invalid agents: claude and lists claude-code as the valid name. Use -a claude-code.

Landing path differs by adapter: claude-code writes to its own .claude/skills/; codex and cursor both write to the shared .agents/skills/; pi writes to .pi/skills/. When you classify or audit, .claude/, .agents/, and .pi/ are all Class E (tool/agent state — keep ignored). Path is what matters, not adapter.

Limits