@esauflores/pi-agent-prompts0.1.1
Coding profiles: inject a prompt file as a persistent system-prompt mode
Untested on omp
This is a pi package. No one has reported whether it works on omp yet. If you try it, add your result below.
Reports 0
No reports yet.
Readme
@esauflores/pi-agent-prompts
Turn any markdown file in prompts/ into a slash command that injects it as a persistent coding
profile.
/coding-full activate (press again to switch off)
/coding-plan
/coding-review
/coding-solo
The profile rides in a system-prompt section, so it outranks turn history: it applies to every
request until you toggle it off, instead of being a one-shot message that later turns erode. The
active profile shows in the footer as ◈ coding-solo.
Install
pi install npm:@esauflores/pi-agent-prompts
Add your own profiles
Drop a .md file into prompts/ and run /reload. The filename becomes the command name:
prompts/
├── coding-full.md → /coding-full
├── coding-plan.md → /coding-plan
├── coding-review.md → /coding-review
└── coding-solo.md → /coding-solo
Add foo.md, get /foo — no registration, no config. The first heading or line of the file becomes
the command description in the menu. Names must be
[a-z][a-z0-9-]*; other files are ignored.
Install into a project instead
To keep profiles with a repository, copy the extension into the project and point at it:
{ "packages": ["./.pi/pi-agent-prompts"] }
Behavior
- Running the active profile's command again turns it off.
- The choice is stored as a branch-local session entry, so
/treenavigation and forks see the profile belonging to that branch, not the last one you picked. - It survives
/reloadand session resume; the state is restored on bothsession_startandsession_tree.
Limits
- Profiles are added to the system prompt, so a large one costs context on every request.
- Only the newest profile entry on the current branch is active; activating another replaces it rather than stacking.