# oh-my-copilot

Default behaviours installed with this repo. Override per project as needed.

## Approach
- Surface assumptions before coding.
- Prefer the simplest change that satisfies the request.
- Touch only what the task requires.
- Verify success with concrete checks: tests, output, behaviour.

## Validation
- Run tests for code you change.
- Read the diff before committing.
- If unsure about scope, ask.

## Cost/token discipline
Cost data is local, best-effort, and estimated. `omp cost [--today] [--session <id>]`
summarizes prompt/tool token estimates from the hook ledger; it is not provider billing.

The cost hooks apply when this plugin's `hooks/hooks.json` is active in a Copilot CLI
session. They give session-wide visibility for skills invoked inside that session, not
standalone coverage for copied skills, raw shell scripts, or external CLIs.

Before rerunning noisy commands or failed edits, inspect the latest output and narrow the
next attempt. Prefer bounded summaries for large logs. Oversized postToolUse output is
minimized before it re-enters model context, with raw output preserved on disk and savings
recorded in the cost ledger. Diagnostics (errors, stack traces, assertions) are preserved
inline; other trimmed detail must be recovered by re-reading the raw file path noted in the
hook output (an extra tool call), so full fidelity depends on the model following that pointer.
Budget gates and retry-cost guidance are not current live behavior.

## Skills
Slash commands under `.github/skills/<name>/SKILL.md` are auto-discovered by Copilot. See `omp list` for the catalog active in this project.

## Hooks
Lifecycle hooks declared in `hooks/hooks.json` invoke scripts in `scripts/`. Run `omp doctor` to verify discovery.

## Session lifecycle: self-evolve

Before ending a session, invoke `/self-evolve` to capture mistake patterns.

Trigger when any of these apply:

- The user says "done", "bye", "thanks", "wrapping up", "ship it", or types `/exit`.
- You finish the user's last task with no pending todos and they have not requested follow-up.

Skip when the session was fewer than 3 turns or contained no user corrections.

Invoke `/self-evolve` **at most once per session.** After it reports back, do not invoke it again — even if the user replies "thanks" or "done" to the report itself. Just acknowledge and end the session.

The skill lives at `.github/skills/self-evolve/SKILL.md`. It is self-contained: it tells you which signals count as corrections, where to log them, and when a recurring pattern should become a draft skill.

## Daily log & project goal

Keep lightweight memory in `.omp/` so the next session has context. **Be sparse** — these are
memory aids, not a transcript. Float kills their value.

- **Project goal** (`.omp/goal.md`, via `/project-goal`): set/update the one-line objective when it is
  established or genuinely shifts. Not every session.
- **Session Goal** (via `/goal`): one durable unfinished objective for the current
  Copilot session. AgentStop owns its turn counter; the agent records completion
  and 20-turn extension decisions with idempotent `omp goal complete|extend`
  commands before the stop hook.
- **Daily log** (`.omp/memory/daily/<date>.md`, via `/daily-log`):
  - When a heavy mode (`/ralph`, `/ralplan`, `/omp-autopilot`, `/team`, `/ultrawork`,
    `/ultraqa`) starts, set today's goal from its objective: `omp daily-log set-goal "<text>"`.
  - When such a mode finishes, add **one** concise summary entry — `omp daily-log add "<text>"`:
    what changed, key decisions, next step.
  - Otherwise add an entry only at a genuine milestone you judge worth remembering.
- **Project memory**: `omp project-memory add-directive "<rule>"` for a must-follow rule —
  directives are **injected at every session start** (never on-demand), so add only real,
  lasting rules and keep them few. `omp project-memory add-note "<title>" [--body "<text>"]`
  for durable facts; list them with `omp project-memory index` and load one with
  `omp project-memory read <id>` — bodies stay on disk until asked, so notes never bloat context.
- **Caps & retention:** keep directives to a handful (only the first ~12 are injected at start);
  a few daily entries per day — decisions and outcomes, not narration; skip trivial sessions;
  never log per skill invocation. Trim old day-files with `omp daily-log prune [--keep-days <n>]`
  (default 30).