AI Agent Skills

pass-cli ships an agent usage guide inside the binary. Any AI agent (Claude Code, Cursor, Codex, …) driving pass-cli can load version-matched guidance straight from the CLI, so the instructions never drift from the installed version.

For agents: start here

pass-cli skills get core          # safe-usage guide: exec/export/inject/agent/list/get + leak traps
pass-cli skills get core --full   # also include the full command reference
pass-cli skills list              # list every skill shipped with this version

skills get core is the entry point. It explains the one rule that matters — never let a secret value land in the transcript, a log, or a watched file — and which command to reach for (exec by default; inject for composite secrets; list -q to discover services; get only as a last resort).

Installing the discovery stub

skills install drops a small skill file into your agent’s skills directory so the agent discovers pass-cli automatically and is pointed at skills get core:

pass-cli skills install                 # auto-detects ~/.claude/skills or ~/.agents/skills
pass-cli skills install --dir <path>    # choose a specific directory
pass-cli skills install --force         # overwrite an existing, differing stub

The stub is intentionally thin — the real, version-matched guidance stays in the binary. Re-running skills install after upgrading pass-cli is safe (it’s a no-op when the stub is already current).

Why embedded, not a static doc

The skill content is compiled into the binary via //go:embed, so it’s always in lockstep with the commands and flags that binary actually has. A stale copy pasted into a project’s docs or a user’s dotfiles can’t drift out of sync, because agents read the guide from the CLI itself.