agentsclimarketplace

Op

Skill aarshsharma2803-stack/optimusprime/skills/op

The session state protocol for AI coding. Hook-level scope enforcement, cross-session memory, contradiction detection, and predictive context - Claude is stateless, your project isn't.

Install
npx -y skills add aarshsharma2803-stack/optimusprime --skill op

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

OptimusPrime — short trigger. Identical to /optimusprime. Session memory, guardrails, output compression, Auto Bots. Subcommands: status | compact | slim | repair | bots | tokens | quality | dashboard Trigger: /op, "op status", "op slim", "what does op say", "check op", "op repair"

SKILL.md

3.2 KB, as published. Nobody here has run it

/op = /optimusprime. Same skill, shorter to type.

All data lives in .optimusprime/ — walked up from cwd, auto-created on first run.

/op (no args) — activate

1. Detect session type. Read .optimusprime/session-state.jsonprompt_count. Over 5 = existing; otherwise new.

New session:

⚡ OP ACTIVE
Caveman Bot on · compression on · guardrails armed
Modes: status | compact | repair | bots | tokens | quality | dashboard
What are you working on?

Existing session: recover BEFORE anything else. Read contract.json (goal), session-snapshot.md (last state), decisions.md (last 5), attempts.md (never retry these), loop-state.json. Show:

⚡ CONTEXT RECOVERED
Goal: <goal> · <n> decisions · <n>k tokens · loop <n>
Avoid: <failed approaches>
Next: <next action from snapshot>

Both flows: output-mode compression active from this response onward.

/op status

Tokens + cost + decisions + loop streak + compression % + active bots. One panel, nothing else.

/op compact

Compact now. Keep: goal, decisions+reasons, errors+context, code written, failed approaches. Drop: repeated explanations, narration, filler. Write to .optimusprime/session-snapshot.md.

/op slim

Maximum token efficiency — does two things at once:

1. Compact conversation (same as /op compact — strip history, keep only what matters).

2. Escalate output to ultra-terse for the rest of this session:

  • Code tasks: code block only, zero sentences after
  • Questions: one word or one line max
  • No preamble ever. No wrap-up ever. No confirmation sentences.
  • Caveman ultra mode: fragments, no articles, arrows for causality

Show before/after token estimate and confirm: ⚡ SLIM MODE — <n>k tokens saved · ultra-terse until session end

Use when: tokens are high and you need maximum throughput with minimum waste.

/op repair

Read loop-state.json + attempts.md. Show current error + what failed. Propose ONE different approach. Reset loop-state.json.

/op bots

Table of 5 bots: Caveman (tokens>40k), Superpowers (full builds), UI/UX Pro Max (frontend), Ponytail (minimal budget), Gstack (deploy). Offer mode change.

/op tokens

Session tokens + cost + last 5 trend + compression savings. Threshold guide: <40k healthy · 40-80k caveman on · 80k+ compact · 150k+ new session.

/op quality

Scan changed files: functions >30 lines, duplicate logic, hardcoded secrets, SQL concat, eval on user input. Critical / important / minor. ✅ clean if nothing found.

/op dashboard

status + last 5 decisions + open TODOs + next action, one screen.

Auto rules

  • tokens > 40k → caveman full auto
  • tokens > 80k → max compression + suggest compact
  • loop ≥ 3 → suggest repair
  • frontend files → UI/UX rigor
  • minimal budget → ponytail (smallest correct impl)

Off: "stop op" / "deactivate op" / "stop optimusprime".

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.