Pointer-file mode

If you'd rather have each agent read tastecode.md on its own — without going through tastecode use every time — run:

tastecode install --all

This drops a small pointer file into each agent's native config format, telling that agent to read tastecode.md before coding:

AgentFile written
Claude CodeCLAUDE.md + .claude/commands/tastecode-*.md
Cursor.cursor/rules/tastecode.mdc
Cline.clinerules/tastecode.md
Codex CLIAGENTS.md
OpenCodeAGENTS.md (shared with Codex)
AiderCONVENTIONS.md

Each pointer is wrapped in fence comments (<!-- tastecode:start --> ... <!-- tastecode:end -->) so re-running install updates in place — never duplicates.

Claude Code slash commands

For Claude Code specifically, install also drops slash-command files under .claude/commands/, so TasteCode is reachable from inside the Claude REPL:

Slash commandWhat it runs
/tastecode-reject <reason>tastecode reject "<reason>"
/tastecode-accept <note>tastecode accept "<note>"
/tastecode-learntastecode learn
/tastecode-inittastecode init
/tastecode-doctortastecode doctor
/tastecode-providerstastecode providers

These show up in Claude Code's slash-command menu the next time you open the project. Each is a thin shell-out to the tastecode binary — make sure it's on your PATH (any global or npx install works).